mirrors
/
PyRetri
mirror of https://github.com/PyRetri/PyRetri.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

upload

pull/1/head
hubenyi 2020-04-17 20:31:26 +08:00
parent 402cb98b3b
commit 06afa9b644
23 changed files with 325 additions and 563 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
docs/GETTING_STARTED.md
View File

@ -67,7 +67,7 @@ Examples:
# for dataset collecting images with the same label in one directory # for dataset collecting images with the same label in one directory
python3 main/make_data_json.py -d /data/caltech101/gallery/ -sp data_jsons/caltech_gallery.json -t general python3 main/make_data_json.py -d /data/caltech101/gallery/ -sp data_jsons/caltech_gallery.json -t general
python3 main/make_data_json.py -d /data/caltech101/query/ -sp data_jsons/caltech_query.json -t feneral python3 main/make_data_json.py -d /data/caltech101/query/ -sp data_jsons/caltech_query.json -t general
# for oxford/paris dataset # for oxford/paris dataset
python3 main/make_data_json.py -d /data/cbir/oxford/gallery/ -sp data_jsons/oxford_gallery.json -t oxford -gt /data/cbir/oxford/gt/ python3 main/make_data_json.py -d /data/cbir/oxford/gallery/ -sp data_jsons/oxford_gallery.json -t oxford -gt /data/cbir/oxford/gt/
@ -229,10 +229,10 @@ cd search/
### Define Search Space ### Define Search Space
We decompose the search space into three sub search spaces: data_process, extract and index, each of which corresponds to a specified file. Search space is defined by adding methods with hyper-parameters to a specified dict. You can add a search operator as follows: We decompose the search space into three sub search spaces: pre_process, extract and index, each of which corresponds to a specified file. Search space is defined by adding methods with hyper-parameters to a specified dict. You can add a search operator as follows:
```shell ```shell
data_processes.add( pre_processes.add(
"PadResize224", "PadResize224",
{ {
"batch_size": 32, "batch_size": 32,
@ -257,7 +257,7 @@ data_processes.add(
) )
``` ```
By doing this, a data_process operator named "PadResize224" is added to the data_process sub search space and will be searched in the following process. By doing this, a pre_process operator named "PadResize224" is added to the data_process sub search space and will be searched in the following process.
### Search ### Search
@ -287,7 +287,7 @@ python3 search_extract.py -sp /data/features/gap_gmp_gem_crow_spoc/ -sm search_m
Search for the indexing combinations by: Search for the indexing combinations by:
```shell ```shell
python3 search_query.py [-fd ${fea_dir}] [-sm ${search_modules}] [-sp ${save_path}] python3 search_index.py [-fd ${fea_dir}] [-sm ${search_modules}] [-sp ${save_path}]
``` ```
Arguments: Arguments:
@ -299,8 +299,23 @@ Arguments:
Examples: Examples:
```shell ```shell
python3 search_query.py -fd /data/features/gap_gmp_gem_crow_spoc/ -sm search_modules -sp /data/features/gap_gmp_gem_crow_spoc_result.json python3 search_index.py -fd /data/features/gap_gmp_gem_crow_spoc/ -sm search_modules -sp /data/features/gap_gmp_gem_crow_spoc_result.json
``` ```
#### show search results
We provide two ways to show the search results. One is save all the search results in a csv format file, which can be used for further analyses. The other is showing the search results according to the given keywords. You can define the keywords as follows:
```sh
keywords = {
'data_name': ['market'],
'pre_process_name': list(),
'model_name': list(),
'feature_map_name': list(),
'aggregator_name': list(),
'post_process_name': ['no_fea_process', 'l2_normalize', 'pca_whiten', 'pca_wo_whiten'],
}
```
See show_search_results.py for more details.

2
docs/INSTALL.md
View File

@ -35,7 +35,7 @@ cd pyretri
3. Install PyRetri. 3. Install PyRetri.
```shell ```shell
python setup.py install python3 setup.py install
``` ```
## Prepare Datasets ## Prepare Datasets

2
docs/MODEL_ZOO.md
View File

@ -35,7 +35,7 @@ Choosing the implementations mentioned above as baselines and adding some tricks
For person re-identification, we use the model provided by [Person_reID_baseline](https://github.com/layumi/Person_reID_baseline_pytorch) and reproduce its resutls. In addition, we train a model on DukeMTMC-reID through the open source code for further experiments. For person re-identification, we use the model provided by [Person_reID_baseline](https://github.com/layumi/Person_reID_baseline_pytorch) and reproduce its resutls. In addition, we train a model on DukeMTMC-reID through the open source code for further experiments.
###pre-trained models ### pre-trained models
| Training Set | Backbone | for Short | Download | | Training Set | Backbone | for Short | Download |
| :-----------: | :-------: | :-------: | :------: | | :-----------: | :-------: | :-------: | :------: |

4
pyretri/index/dim_processor/dim_processors_impl/part_pca.py
View File

@ -84,7 +84,9 @@ class PartPCA(DimProcessorBase):
pca = self.pcas[fea_name]["pca"] pca = self.pcas[fea_name]["pca"]
ori_fea = fea[:, st_idx: ed_idx] ori_fea = fea[:, st_idx: ed_idx]
proj_fea = pca.transform(ori_fea) proj_fea = normalize(ori_fea, norm='l2')
proj_fea = pca.transform(proj_fea)
proj_fea = normalize(proj_fea, norm='l2')
ret.append(proj_fea) ret.append(proj_fea)

14
pyretri/index/dim_processor/dim_processors_impl/part_svd.py
View File

@ -65,6 +65,7 @@ class PartSVD(DimProcessorBase):
else: else:
proj_part_dim = self._hyper_params["proj_dim"] - already_proj_dim proj_part_dim = self._hyper_params["proj_dim"] - already_proj_dim
assert proj_part_dim < ori_part_dim, "reduction dimension can not be distributed to each part!" assert proj_part_dim < ori_part_dim, "reduction dimension can not be distributed to each part!"
already_proj_dim += proj_part_dim
svd = SKSVD(n_components=proj_part_dim) svd = SKSVD(n_components=proj_part_dim)
train_fea = fea[:, st_idx: ed_idx] train_fea = fea[:, st_idx: ed_idx]
@ -79,22 +80,23 @@ class PartSVD(DimProcessorBase):
} }
def __call__(self, fea: np.ndarray) -> np.ndarray: def __call__(self, fea: np.ndarray) -> np.ndarray:
if self._hyper_params["proj_dim"] != 0: fea_names = np.sort(list(self.svds.keys()))
ret = np.zeros(shape=(fea.shape[0], self._hyper_params["proj_dim"])) ret = list()
else:
ret = np.zeros(shape=(fea.shape[0], fea.shape[1] - len(self.svds)))
for fea_name in self.svds: for fea_name in fea_names:
st_idx, ed_idx = self.svds[fea_name]["pos"][0], self.svds[fea_name]["pos"][1] st_idx, ed_idx = self.svds[fea_name]["pos"][0], self.svds[fea_name]["pos"][1]
svd = self.svds[fea_name]["svd"] svd = self.svds[fea_name]["svd"]
proj_fea = fea[:, st_idx: ed_idx] proj_fea = fea[:, st_idx: ed_idx]
proj_fea = normalize(proj_fea, norm='l2')
proj_fea = svd.transform(proj_fea) proj_fea = svd.transform(proj_fea)
if self._hyper_params["whiten"]: if self._hyper_params["whiten"]:
proj_fea = proj_fea / (self.svds[fea_name]["std"] + 1e-6) proj_fea = proj_fea / (self.svds[fea_name]["std"] + 1e-6)
proj_fea = normalize(proj_fea, norm='l2')
ret[:, st_idx: ed_idx] = proj_fea ret.append(proj_fea)
ret = np.concatenate(ret, axis=1)
return ret return ret

13
search/reid_search_extract.py
View File

@ -49,24 +49,27 @@ def main():
# load search space # load search space
datasets = load_datasets() datasets = load_datasets()
data_processes = importlib.import_module("{}.data_process_dict".format(args.search_modules)).data_processes pre_processes = importlib.import_module("{}.pre_process_dict".format(args.search_modules)).pre_processes
models = importlib.import_module("{}.extract_dict".format(args.search_modules)).models models = importlib.import_module("{}.extract_dict".format(args.search_modules)).models
extracts = importlib.import_module("{}.extract_dict".format(args.search_modules)).extracts extracts = importlib.import_module("{}.extract_dict".format(args.search_modules)).extracts
# search in an exhaustive way # search in an exhaustive way
for data_name, data_args in datasets.items(): for data_name, data_args in datasets.items():
for data_proc_name, data_proc_args in data_processes.items(): for pre_proc_name, pre_proc_args in pre_processes.items():
if 'market' in data_name: if 'market' in data_name:
model_name = 'market_res50' model_name = 'market_res50'
elif 'duke' in data_name: elif 'duke' in data_name:
model_name = 'duke_res50' model_name = 'duke_res50'
feature_full_name = data_name + "_" + data_proc_name + "_" + model_name feature_full_name = data_name + "_" + pre_proc_name + "_" + model_name
print(feature_full_name) print(feature_full_name)
if os.path.exists(os.path.join(args.save_path, feature_full_name)):
print("[Search Extract]: config exists...")
continue
# load retrieval pipeline settings # load retrieval pipeline settings
cfg.datasets.merge_from_other_cfg(data_proc_args) cfg.datasets.merge_from_other_cfg(pre_proc_args)
cfg.model.merge_from_other_cfg(models[model_name]) cfg.model.merge_from_other_cfg(models[model_name])
cfg.extract.merge_from_other_cfg(extracts[model_name]) cfg.extract.merge_from_other_cfg(extracts[model_name])

29
search/new_reid_search_query.py → search/reid_search_index.py
View File

@ -5,7 +5,7 @@ import importlib
import os import os
import argparse import argparse
from .utils.misc import check_exist, get_dir from utils.misc import check_result_exist, get_dir, get_default_result_dict
from pyretri.config import get_defaults_cfg from pyretri.config import get_defaults_cfg
from pyretri.index import build_index_helper, feature_loader from pyretri.index import build_index_helper, feature_loader
@ -53,8 +53,8 @@ def main():
# load search space # load search space
datasets = load_datasets() datasets = load_datasets()
queries = importlib.import_module("{}.query_dict".format(args.search_modules)).queries indexes = importlib.import_module("{}.index_dict".format(args.search_modules)).indexes
evaluates = importlib.import_module("{}.query_dict".format(args.search_modules)).evaluates evaluates = importlib.import_module("{}.index_dict".format(args.search_modules)).evaluates
if os.path.exists(args.save_path): if os.path.exists(args.save_path):
with open(args.save_path, "r") as f: with open(args.save_path, "r") as f:
@ -64,8 +64,9 @@ def main():
for dir in os.listdir(args.fea_dir): for dir in os.listdir(args.fea_dir):
for data_name, data_args in datasets.items(): for data_name, data_args in datasets.items():
for query_name, query_args in queries.items(): for index_name, index_args in indexes.items():
if data_name in dir: if data_name in dir:
print(dir)
# get dirs # get dirs
gallery_fea_dir, query_fea_dir, train_fea_dir = get_dir(args.fea_dir, dir, data_args) gallery_fea_dir, query_fea_dir, train_fea_dir = get_dir(args.fea_dir, dir, data_args)
@ -73,21 +74,19 @@ def main():
# get evaluate setting # get evaluate setting
evaluate_args = evaluates["reid_overall"] evaluate_args = evaluates["reid_overall"]
for dim_proc in query_args.dim_processors.names: for dim_proc in index_args.dim_processors.names:
if dim_proc in ["PartPCA", "PartSVD", "PCA", "SVD"]: if dim_proc in ["PartPCA", "PartSVD", "PCA", "SVD"]:
query_args.dim_processors[dim_proc].train_fea_dir = train_fea_dir index_args.dim_processors[dim_proc].train_fea_dir = train_fea_dir
for fea_name in fea_names: for fea_name in fea_names:
result_dict = get_default_result_dict(dir, data_name, index_name, fea_name)
result_dict = get_default_result_dict(dir, data_name, query_name, fea_name) if check_result_exist(result_dict, results):
if check_exist(result_dict, results):
print("[Search Query]: config exists...") print("[Search Query]: config exists...")
continue continue
print(data_name + '_' + fea_name + '_' + query_name)
# load retrieval pipeline settings # load retrieval pipeline settings
query_args.feature_names = [fea_name] index_args.feature_names = [fea_name]
cfg.index.merge_from_other_cfg(query_args) cfg.index.merge_from_other_cfg(index_args)
cfg.evaluate.merge_from_other_cfg(evaluate_args) cfg.evaluate.merge_from_other_cfg(evaluate_args)
# load features # load features
@ -95,12 +94,12 @@ def main():
gallery_fea, gallery_info, _ = feature_loader.load(gallery_fea_dir, [fea_name]) gallery_fea, gallery_info, _ = feature_loader.load(gallery_fea_dir, [fea_name])
# build helper and index features # build helper and index features
query_helper = build_index_helper(cfg.index) index_helper = build_index_helper(cfg.index)
query_result_info, _, _ = query_helper.do_index(query_fea, query_info, gallery_fea) index_result_info, _, _ = index_helper.do_index(query_fea, query_info, gallery_fea)
# build helper and evaluate results # build helper and evaluate results
evaluate_helper = build_evaluate_helper(cfg.evaluate) evaluate_helper = build_evaluate_helper(cfg.evaluate)
mAP, recall_at_k = evaluate_helper.do_eval(query_result_info, gallery_info) mAP, recall_at_k = evaluate_helper.do_eval(index_result_info, gallery_info)
# record results # record results
to_save_recall = dict() to_save_recall = dict()

4
search/reid_search_modules/extract_dict.py
View File

@ -19,6 +19,7 @@ models.add(
extracts.add( extracts.add(
"market_res50", "market_res50",
{ {
"assemble": 1,
"extractor": { "extractor": {
"name": "ReIDSeries", "name": "ReIDSeries",
"ReIDSeries": { "ReIDSeries": {
@ -39,13 +40,14 @@ models.add(
{ {
"name": "ft_net", "name": "ft_net",
"ft_net": { "ft_net": {
"load_checkpoint": "/home/songrenjie/projects/reID_baseline/model/ft_ResNet50/net_59.pth" "load_checkpoint": "/home/songrenjie/projects/reID_baseline/model/ft_ResNet50/res50_duke.pth"
} }
} }
) )
extracts.add( extracts.add(
"duke_res50", "duke_res50",
{ {
"assemble": 1,
"extractor": { "extractor": {
"name": "ReIDSeries", "name": "ReIDSeries",
"ReIDSeries": { "ReIDSeries": {

44
search/reid_search_modules/query_dict.py → search/reid_search_modules/index_dict.py
View File

@ -3,10 +3,10 @@
from utils.search_modules import SearchModules from utils.search_modules import SearchModules
from pyretri.config import get_defaults_cfg from pyretri.config import get_defaults_cfg
queries = SearchModules() indexes = SearchModules()
evaluates = SearchModules() evaluates = SearchModules()
queries.add( indexes.add(
"no_fea_process", "no_fea_process",
{ {
"gallery_fea_dir": "", "gallery_fea_dir": "",
@ -32,7 +32,7 @@ queries.add(
} }
) )
queries.add( indexes.add(
"l2_normalize", "l2_normalize",
{ {
"gallery_fea_dir": "", "gallery_fea_dir": "",
@ -58,7 +58,7 @@ queries.add(
} }
) )
queries.add( indexes.add(
"pca_wo_whiten", "pca_wo_whiten",
{ {
"gallery_fea_dir": "", "gallery_fea_dir": "",
@ -67,11 +67,12 @@ queries.add(
"feature_names": [], "feature_names": [],
"dim_processors": { "dim_processors": {
"names": ["PartPCA"], "names": ["L2Normalize", "PCA", "L2Normalize"],
"PartPCA": { "PCA": {
"whiten": False, "whiten": False,
"train_fea_dir": "", "train_fea_dir": "",
"proj_dim": 512 "proj_dim": 512,
"l2": True,
} }
}, },
@ -89,7 +90,7 @@ queries.add(
} }
) )
queries.add( indexes.add(
"pca_whiten", "pca_whiten",
{ {
"gallery_fea_dir": "", "gallery_fea_dir": "",
@ -98,11 +99,12 @@ queries.add(
"feature_names": [], "feature_names": [],
"dim_processors": { "dim_processors": {
"names": ["PartPCA"], "names": ["L2Normalize", "PCA", "L2Normalize"],
"PartPCA": { "PCA": {
"whiten": True, "whiten": True,
"train_fea_dir": "", "train_fea_dir": "",
"proj_dim": 512 "proj_dim": 512,
"l2": True,
} }
}, },
@ -120,7 +122,7 @@ queries.add(
} }
) )
queries.add( indexes.add(
"svd_wo_whiten", "svd_wo_whiten",
{ {
"gallery_fea_dir": "", "gallery_fea_dir": "",
@ -129,11 +131,12 @@ queries.add(
"feature_names": [], "feature_names": [],
"dim_processors": { "dim_processors": {
"names": ["PartSVD"], "names": ["L2Normalize", "SVD", "L2Normalize"],
"PartSVD": { "SVD": {
"whiten": False, "whiten": False,
"train_fea_dir": "", "train_fea_dir": "",
"proj_dim": 511 "proj_dim": 511,
"l2": True,
} }
}, },
@ -151,7 +154,7 @@ queries.add(
} }
) )
queries.add( indexes.add(
"svd_whiten", "svd_whiten",
{ {
"gallery_fea_dir": "", "gallery_fea_dir": "",
@ -160,11 +163,12 @@ queries.add(
"feature_names": [], "feature_names": [],
"dim_processors": { "dim_processors": {
"names": ["PartSVD"], "names": ["L2Normalize", "SVD", "L2Normalize"],
"PartSVD": { "SVD": {
"whiten": True, "whiten": True,
"train_fea_dir": "", "train_fea_dir": "",
"proj_dim": 511 "proj_dim": 511,
"l2": True,
} }
}, },
@ -211,5 +215,5 @@ evaluates.add(
cfg = get_defaults_cfg() cfg = get_defaults_cfg()
queries.check_valid(cfg["index"]) indexes.check_valid(cfg["index"])
evaluates.check_valid(cfg["evaluate"]) evaluates.check_valid(cfg["evaluate"])

6
search/reid_search_modules/data_process_dict.py → search/reid_search_modules/pre_process_dict.py
View File

@ -3,9 +3,9 @@
from utils.search_modules import SearchModules from utils.search_modules import SearchModules
from pyretri.config import get_defaults_cfg from pyretri.config import get_defaults_cfg
data_processes = SearchModules() pre_processes = SearchModules()
data_processes.add( pre_processes.add(
"Direct256128", "Direct256128",
{ {
"batch_size": 32, "batch_size": 32,
@ -31,4 +31,4 @@ data_processes.add(
cfg = get_defaults_cfg() cfg = get_defaults_cfg()
data_processes.check_valid(cfg["datasets"]) pre_processes.check_valid(cfg["datasets"])

149
search/reid_search_query.py
View File

@ -1,149 +0,0 @@
# -*- coding: utf-8 -*-
import json
import importlib
import os
import argparse
from pyretri.config import get_defaults_cfg
from pyretri.query import build_query_helper
from pyretri.evaluate import build_evaluate_helper
from pyretri.query.utils import feature_loader
fea_names = ["output"]
task_mapping = {
"market_gallery": {
"gallery": "market_gallery",
"query": "market_query",
"train_fea_dir": "market_gallery"
},
"duke_gallery": {
"gallery": "duke_gallery",
"query": "duke_query",
"train_fea_dir": "duke_gallery"
},
}
def check_exist(now_res, exist_results):
for e_r in exist_results:
totoal_equal = True
for key in now_res:
if now_res[key] != e_r[key]:
totoal_equal = False
break
if totoal_equal:
return True
return False
def parse_args():
parser = argparse.ArgumentParser(description='A tool box for deep learning-based image retrieval')
parser.add_argument('--fea_dir', '-f', default=None, type=str, help="path of feature directory", required=True)
parser.add_argument(
"--search_modules",
"-m",
default="",
help="name of search module's directory",
type=str,
required=True
)
parser.add_argument("--save_path", "-s", default=None, type=str, required=True)
args = parser.parse_args()
return args
def get_default_result_dict(dir, task_name, query_name, fea_name):
result_dict = {
"task_name": task_name.split("_")[0],
"dataprocess": dir.split("_")[0],
"model_name": "_".join(dir.split("_")[-2:]),
"feature_map_name": fea_name.split("_")[0],
"fea_process_name": query_name
}
if fea_name == "output":
result_dict["aggregator_name"] = "none"
else:
result_dict["aggregator_name"] = fea_name.split("_")[1]
return result_dict
def main():
args = parse_args()
cfg = get_defaults_cfg()
queries = importlib.import_module("{}.query_dict".format(args.search_modules)).queries
evaluates = importlib.import_module("{}.query_dict".format(args.search_modules)).evaluates
if os.path.exists(args.save_path):
with open(args.save_path, "r") as f:
results = json.load(f)
else:
results = list()
q_cnt = 0
for dir in os.listdir(args.fea_dir):
q_cnt += 1
print("Processing {} / {} queries...".format(q_cnt, len(queries)))
for task_name in task_mapping:
for query_name, query_args in queries.items():
if task_name in dir:
for fea_name in fea_names:
gallery_fea_dir = os.path.join(args.fea_dir, dir)
query_fea_dir = gallery_fea_dir.replace(task_name, task_mapping[task_name]["query"])
train_fea_dir = gallery_fea_dir.replace(task_name, task_mapping[task_name]["train_fea_dir"])
assert os.path.exists(gallery_fea_dir), gallery_fea_dir
assert os.path.exists(query_fea_dir), query_fea_dir
assert os.path.exists(train_fea_dir), train_fea_dir
for post_proc in ["PartPCA", "PartSVD"]:
if post_proc in query_args.post_processors.names:
query_args.post_processors[post_proc].train_fea_dir = train_fea_dir
query_args.gallery_fea_dir, query_args.query_fea_dir = gallery_fea_dir, query_fea_dir
query_args.feature_names = [fea_name]
eval_args = evaluates["reid_overall"]
result_dict = get_default_result_dict(dir, task_name, query_name, fea_name)
if check_exist(result_dict, results):
print("[Search Query]: config exists...")
continue
cfg.query.merge_from_other_cfg(query_args)
cfg.evaluate.merge_from_other_cfg(eval_args)
query_helper = build_query_helper(cfg.query)
evaluate_helper = build_evaluate_helper(cfg.evaluate)
query_fea, query_info_dicts, _ = feature_loader.load(cfg.query.query_fea_dir,
cfg.query.feature_names)
gallery_fea, gallery_info_dicts, _ = feature_loader.load(cfg.query.gallery_fea_dir,
cfg.query.feature_names)
query_result_info_dicts, _, _ = query_helper.do_query(query_fea, query_info_dicts, gallery_fea)
mAP, recall_at_k = evaluate_helper.do_eval(query_result_info_dicts, gallery_info_dicts)
to_save_dict = dict()
for k in recall_at_k:
to_save_dict[str(k)] = recall_at_k[k]
result_dict["mAP"] = float(mAP)
result_dict["recall_at_k"] = to_save_dict
print(result_dict)
assert False
results.append(result_dict)
with open(args.save_path, "w") as f:
json.dump(results, f)
if __name__ == '__main__':
main()

18
search/search_extract.py
View File

@ -30,17 +30,11 @@ def load_datasets():
return datasets return datasets
def check_exist(save_path, full_name):
if os.path.exists(os.path.join(save_path, full_name)):
return True
return False
def parse_args(): def parse_args():
parser = argparse.ArgumentParser(description='A tool box for deep learning-based image retrieval') parser = argparse.ArgumentParser(description='A tool box for deep learning-based image retrieval')
parser.add_argument('opts', default=None, nargs=argparse.REMAINDER) parser.add_argument('opts', default=None, nargs=argparse.REMAINDER)
parser.add_argument('--save_path', '-sp', default=None, type=str, help="save path for feature") parser.add_argument('--save_path', '-sp', default=None, type=str, help="save path for feature")
parser.add_argument("--search_modules", "-sm", default="", type=str, help="name of search module's directory") parser.add_argument("--search_modules", "-sm", default=None, type=str, help="name of search module's directory")
args = parser.parse_args() args = parser.parse_args()
return args return args
@ -58,24 +52,24 @@ def main():
# load search space # load search space
datasets = load_datasets() datasets = load_datasets()
data_processes = importlib.import_module("{}.data_process_dict".format(args.search_modules)).data_processes pre_processes = importlib.import_module("{}.pre_process_dict".format(args.search_modules)).pre_processes
models = importlib.import_module("{}.extract_dict".format(args.search_modules)).models models = importlib.import_module("{}.extract_dict".format(args.search_modules)).models
extracts = importlib.import_module("{}.extract_dict".format(args.search_modules)).extracts extracts = importlib.import_module("{}.extract_dict".format(args.search_modules)).extracts
# search in an exhaustive way # search in an exhaustive way
for data_name, data_args in datasets.items(): for data_name, data_args in datasets.items():
for data_proc_name, data_proc_args in data_processes.items(): for pre_proc_name, pre_proc_args in pre_processes.items():
for model_name, model_args in models.items(): for model_name, model_args in models.items():
feature_full_name = data_name + "_" + data_proc_name + "_" + model_name feature_full_name = data_name + "_" + pre_proc_name + "_" + model_name
print(feature_full_name) print(feature_full_name)
if check_exist(args.save_path, feature_full_name): if os.path.exists(os.path.join(args.save_path, feature_full_name)):
print("[Search Extract]: config exists...") print("[Search Extract]: config exists...")
continue continue
# load retrieval pipeline settings # load retrieval pipeline settings
cfg.datasets.merge_from_other_cfg(data_proc_args) cfg.datasets.merge_from_other_cfg(pre_proc_args)
cfg.model.merge_from_other_cfg(model_args) cfg.model.merge_from_other_cfg(model_args)
cfg.extract.merge_from_other_cfg(extracts[model_name]) cfg.extract.merge_from_other_cfg(extracts[model_name])

44
search/new_search_query.py → search/search_index.py
View File

@ -5,13 +5,26 @@ import importlib
import os import os
import argparse import argparse
from .utils.misc import check_exist, get_dir from utils.misc import check_result_exist, get_dir, get_default_result_dict
from pyretri.config import get_defaults_cfg from pyretri.config import get_defaults_cfg
from pyretri.index import build_index_helper, feature_loader from pyretri.index import build_index_helper, feature_loader
from pyretri.evaluate import build_evaluate_helper from pyretri.evaluate import build_evaluate_helper
# # gap, gmp, gem, spoc, crow
# vgg_fea = ["pool4_GAP", "pool4_GMP", "pool4_GeM", "pool4_SPoC", "pool4_Crow",
# "pool5_GAP", "pool5_GMP", "pool5_GeM", "pool5_SPoC", "pool5_Crow",
# "fc"]
# res_fea = ["pool3_GAP", "pool3_GMP", "pool3_GeM", "pool3_SPoC", "pool4_Crow",
# "pool4_GAP", "pool4_GMP", "pool4_GeM", "pool4_SPoC", "pool4_Crow",
# "pool5_GAP", "pool5_GMP", "pool5_GeM", "pool5_SPoC", "pool5_Crow"]
# # scda, rmca
# vgg_fea = ["pool5_SCDA", "pool5_RMAC"]
# res_fea = ["pool5_SCDA", "pool5_RMAC"]
# pwa
vgg_fea = ["pool5_PWA"] vgg_fea = ["pool5_PWA"]
res_fea = ["pool5_PWA"] res_fea = ["pool5_PWA"]
@ -80,8 +93,8 @@ def main():
# load search space # load search space
datasets = load_datasets() datasets = load_datasets()
queries = importlib.import_module("{}.query_dict".format(args.search_modules)).queries indexes = importlib.import_module("{}.index_dict".format(args.search_modules)).indexes
evaluates = importlib.import_module("{}.query_dict".format(args.search_modules)).evaluates evaluates = importlib.import_module("{}.index_dict".format(args.search_modules)).evaluates
if os.path.exists(args.save_path): if os.path.exists(args.save_path):
with open(args.save_path, "r") as f: with open(args.save_path, "r") as f:
@ -91,8 +104,10 @@ def main():
for dir in os.listdir(args.fea_dir): for dir in os.listdir(args.fea_dir):
for data_name, data_args in datasets.items(): for data_name, data_args in datasets.items():
for query_name, query_args in queries.items(): for index_name, index_args in indexes.items():
if data_name in dir: if data_name in dir:
print(dir)
# get dirs # get dirs
gallery_fea_dir, query_fea_dir, train_fea_dir = get_dir(args.fea_dir, dir, data_args) gallery_fea_dir, query_fea_dir, train_fea_dir = get_dir(args.fea_dir, dir, data_args)
@ -102,19 +117,20 @@ def main():
# get feature names # get feature names
fea_names = get_fea_names(gallery_fea_dir) fea_names = get_fea_names(gallery_fea_dir)
for post_proc in query_args.post_processors.names: # set train feature path for dimension reduction processes
if post_proc in ["PartPCA", "PartSVD", "PCA", "SVD"]: for dim_proc in index_args.dim_processors.names:
query_args.post_processors[post_proc].train_fea_dir = train_fea_dir if dim_proc in ["PartPCA", "PartSVD", "PCA", "SVD"]:
index_args.dim_processors[dim_proc].train_fea_dir = train_fea_dir
for fea_name in fea_names: for fea_name in fea_names:
result_dict = get_default_result_dict(dir, data_name, query_name, fea_name) result_dict = get_default_result_dict(dir, data_name, index_name, fea_name)
if check_exist(result_dict, results): if check_result_exist(result_dict, results):
print("[Search Query]: config exists...") print("[Search Query]: config exists...")
continue continue
# load retrieval pipeline settings # load retrieval pipeline settings
query_args.feature_names = [fea_name] index_args.feature_names = [fea_name]
cfg.index.merge_from_other_cfg(query_args) cfg.index.merge_from_other_cfg(index_args)
cfg.evaluate.merge_from_other_cfg(evaluate_args) cfg.evaluate.merge_from_other_cfg(evaluate_args)
# load features # load features
@ -122,12 +138,12 @@ def main():
gallery_fea, gallery_info, _ = feature_loader.load(gallery_fea_dir, [fea_name]) gallery_fea, gallery_info, _ = feature_loader.load(gallery_fea_dir, [fea_name])
# build helper and index features # build helper and index features
query_helper = build_query_helper(cfg.query) index_helper = build_index_helper(cfg.index)
query_result_info, _, _ = query_helper.do_query(query_fea, query_info, gallery_fea) index_result_info, _, _ = index_helper.do_index(query_fea, query_info, gallery_fea)
# build helper and evaluate results # build helper and evaluate results
evaluate_helper = build_evaluate_helper(cfg.evaluate) evaluate_helper = build_evaluate_helper(cfg.evaluate)
mAP, recall_at_k = evaluate_helper.do_eval(query_result_info, gallery_info) mAP, recall_at_k = evaluate_helper.do_eval(index_result_info, gallery_info)
# record results # record results
to_save_recall = dict() to_save_recall = dict()

40
search/search_modules/query_dict.py → search/search_modules/index_dict.py
View File

@ -3,10 +3,10 @@
from utils.search_modules import SearchModules from utils.search_modules import SearchModules
from pyretri.config import get_defaults_cfg from pyretri.config import get_defaults_cfg
queries = SearchModules() indexes = SearchModules()
evaluates = SearchModules() evaluates = SearchModules()
queries.add( indexes.add(
"pca_wo_whiten", "pca_wo_whiten",
{ {
"gallery_fea_dir": "", "gallery_fea_dir": "",
@ -15,11 +15,12 @@ queries.add(
"feature_names": [], "feature_names": [],
"dim_processors": { "dim_processors": {
"names": ["PartPCA"], "names": ["L2Normalize", "PCA", "L2Normalize"],
"PartPCA": { "PCA": {
"whiten": False, "whiten": False,
"train_fea_dir": "", "train_fea_dir": "",
"proj_dim": 512 "proj_dim": 512,
"l2": True,
} }
}, },
@ -37,7 +38,7 @@ queries.add(
} }
) )
queries.add( indexes.add(
"pca_whiten", "pca_whiten",
{ {
"gallery_fea_dir": "", "gallery_fea_dir": "",
@ -46,11 +47,12 @@ queries.add(
"feature_names": [], "feature_names": [],
"dim_processors": { "dim_processors": {
"names": ["PartPCA"], "names": ["L2Normalize", "PCA", "L2Normalize"],
"PartPCA": { "PCA": {
"whiten": True, "whiten": True,
"train_fea_dir": "", "train_fea_dir": "",
"proj_dim": 512 "proj_dim": 512,
"l2": True,
} }
}, },
@ -68,7 +70,7 @@ queries.add(
} }
) )
queries.add( indexes.add(
"svd_wo_whiten", "svd_wo_whiten",
{ {
"gallery_fea_dir": "", "gallery_fea_dir": "",
@ -77,11 +79,12 @@ queries.add(
"feature_names": [], "feature_names": [],
"dim_processors": { "dim_processors": {
"names": ["PartSVD"], "names": ["L2Normalize", "SVD", "L2Normalize"],
"PartSVD": { "SVD": {
"whiten": False, "whiten": False,
"train_fea_dir": "", "train_fea_dir": "",
"proj_dim": 511 "proj_dim": 511,
"l2": True,
} }
}, },
@ -99,7 +102,7 @@ queries.add(
} }
) )
queries.add( indexes.add(
"svd_whiten", "svd_whiten",
{ {
"gallery_fea_dir": "", "gallery_fea_dir": "",
@ -108,11 +111,12 @@ queries.add(
"feature_names": [], "feature_names": [],
"dim_processors": { "dim_processors": {
"names": ["PartSVD"], "names": ["L2Normalize", "SVD", "L2Normalize"],
"PartSVD": { "SVD": {
"whiten": True, "whiten": True,
"train_fea_dir": "", "train_fea_dir": "",
"proj_dim": 511 "proj_dim": 511,
"l2": True,
} }
}, },
@ -150,5 +154,5 @@ evaluates.add(
cfg = get_defaults_cfg() cfg = get_defaults_cfg()
queries.check_valid(cfg["index"]) indexes.check_valid(cfg["index"])
evaluates.check_valid(cfg["evaluate"]) evaluates.check_valid(cfg["evaluate"])

14
search/search_modules/data_process_dict.py → search/search_modules/pre_process_dict.py
View File

@ -3,9 +3,9 @@
from utils.search_modules import SearchModules from utils.search_modules import SearchModules
from pyretri.config import get_defaults_cfg from pyretri.config import get_defaults_cfg
data_processes = SearchModules() pre_processes = SearchModules()
data_processes.add( pre_processes.add(
"Shorter256Center224", "Shorter256Center224",
{ {
"batch_size": 32, "batch_size": 32,
@ -16,8 +16,8 @@ data_processes.add(
"name": "CollateFn" "name": "CollateFn"
}, },
"transformers": { "transformers": {
"names": ["ResizeShorter", "CenterCrop", "ToTensor", "Normalize"], "names": ["ShorterResize", "CenterCrop", "ToTensor", "Normalize"],
"ResizeShorter": { "ShorterResize": {
"size": 256 "size": 256
}, },
"CenterCrop": { "CenterCrop": {
@ -31,7 +31,7 @@ data_processes.add(
} }
) )
data_processes.add( pre_processes.add(
"Direct224", "Direct224",
{ {
"batch_size": 32, "batch_size": 32,
@ -54,7 +54,7 @@ data_processes.add(
} }
) )
data_processes.add( pre_processes.add(
"PadResize224", "PadResize224",
{ {
"batch_size": 32, "batch_size": 32,
@ -80,4 +80,4 @@ data_processes.add(
cfg = get_defaults_cfg() cfg = get_defaults_cfg()
data_processes.check_valid(cfg["datasets"]) pre_processes.check_valid(cfg["datasets"])

25
search/search_pwa_extract.py
View File

@ -21,9 +21,9 @@ def load_datasets():
"cub_query": os.path.join(data_json_dir, "cub_query.json"), "cub_query": os.path.join(data_json_dir, "cub_query.json"),
"indoor_gallery": os.path.join(data_json_dir, "indoor_gallery.json"), "indoor_gallery": os.path.join(data_json_dir, "indoor_gallery.json"),
"indoor_query": os.path.join(data_json_dir, "indoor_query.json"), "indoor_query": os.path.join(data_json_dir, "indoor_query.json"),
"caltech101_gallery": os.path.join(data_json_dir, "caltech101_gallery.json"), "caltech_gallery": os.path.join(data_json_dir, "caltech_gallery.json"),
"caltech101_query": os.path.join(data_json_dir, "caltech101_query.json"), "caltech_query": os.path.join(data_json_dir, "caltech_query.json"),
"paris": os.path.join(data_json_dir, "paris.json"), "paris_all": os.path.join(data_json_dir, "paris.json"),
} }
for data_path in datasets.values(): for data_path in datasets.values():
assert os.path.exists(data_path), "non-exist dataset path {}".format(data_path) assert os.path.exists(data_path), "non-exist dataset path {}".format(data_path)
@ -48,25 +48,30 @@ def main():
# init retrieval pipeline settings # init retrieval pipeline settings
cfg = get_defaults_cfg() cfg = get_defaults_cfg()
data_processes = importlib.import_module("{}.data_process_dict".format(args.search_modules)).data_processes # load search space
datasets = load_datasets()
pre_processes = importlib.import_module("{}.pre_process_dict".format(args.search_modules)).pre_processes
models = importlib.import_module("{}.extract_dict".format(args.search_modules)).models models = importlib.import_module("{}.extract_dict".format(args.search_modules)).models
extracts = importlib.import_module("{}.extract_dict".format(args.search_modules)).extracts extracts = importlib.import_module("{}.extract_dict".format(args.search_modules)).extracts
datasets = load_datasets()
for data_name, data_args in datasets.items(): for data_name, data_args in datasets.items():
for data_proc_name, data_proc_args in data_processes.items(): for pre_proc_name, pre_proc_args in pre_processes.items():
for model_name, model_args in models.items(): for model_name, model_args in models.items():
feature_full_name = data_process_name + "_" + dataset_name + "_" + model_name feature_full_name = data_name + "_" + pre_proc_name + "_" + model_name
print(feature_full_name) print(feature_full_name)
if os.path.exists(os.path.join(args.save_path, feature_full_name)):
print("[Search Extract]: config exists...")
continue
# load retrieval pipeline settings # load retrieval pipeline settings
cfg.datasets.merge_from_other_cfg(data_proc_args) cfg.datasets.merge_from_other_cfg(pre_proc_args)
cfg.model.merge_from_other_cfg(model_args) cfg.model.merge_from_other_cfg(model_args)
cfg.extract.merge_from_other_cfg(extracts[model_name]) cfg.extract.merge_from_other_cfg(extracts[model_name])
pwa_train_fea_dir = os.path.join("/data/my_features/gap_gmp_gem_crow_spoc", feature_full_name) # set train feature path for pwa
pwa_train_fea_dir = os.path.join("/data/features/test_gap_gmp_gem_crow_spoc", feature_full_name)
if "query" in pwa_train_fea_dir: if "query" in pwa_train_fea_dir:
pwa_train_fea_dir.replace("query", "gallery") pwa_train_fea_dir.replace("query", "gallery")
elif "paris" in pwa_train_fea_dir: elif "paris" in pwa_train_fea_dir:

64
search/search_pwa_modules/query_dict.py → search/search_pwa_modules/index_dict.py
View File

@ -3,10 +3,10 @@
from utils.search_modules import SearchModules from utils.search_modules import SearchModules
from pyretri.config import get_defaults_cfg from pyretri.config import get_defaults_cfg
queries = SearchModules() indexes = SearchModules()
evaluates = SearchModules() evaluates = SearchModules()
queries.add( indexes.add(
"pca_wo_whiten", "pca_wo_whiten",
{ {
"gallery_fea_dir": "", "gallery_fea_dir": "",
@ -14,16 +14,17 @@ queries.add(
"feature_names": [], "feature_names": [],
"post_processor": { "dim_processors": {
"name": "PartPCA", "names": ["L2Normalize", "PCA", "L2Normalize"],
"PartPCA": { "PCA": {
"whiten": False, "whiten": False,
"train_fea_dir": "", "train_fea_dir": "",
"proj_dim": 512 "proj_dim": 512,
"l2": True,
} }
}, },
"database_enhance": { "feature_enhancer": {
"name": "Identity" "name": "Identity"
}, },
@ -31,13 +32,13 @@ queries.add(
"name": "KNN" "name": "KNN"
}, },
"re_rank": { "re_ranker": {
"name": "Identity" "name": "Identity"
} }
} }
) )
queries.add( indexes.add(
"pca_whiten", "pca_whiten",
{ {
"gallery_fea_dir": "", "gallery_fea_dir": "",
@ -45,16 +46,17 @@ queries.add(
"feature_names": [], "feature_names": [],
"post_processor": { "dim_processors": {
"name": "PartPCA", "names": ["L2Normalize", "PCA", "L2Normalize"],
"PartPCA": { "PCA": {
"whiten": True, "whiten": True,
"train_fea_dir": "", "train_fea_dir": "",
"proj_dim": 512 "proj_dim": 512,
"l2": True,
} }
}, },
"database_enhance": { "feature_enhancer": {
"name": "Identity" "name": "Identity"
}, },
@ -62,13 +64,13 @@ queries.add(
"name": "KNN" "name": "KNN"
}, },
"re_rank": { "re_ranker": {
"name": "Identity" "name": "Identity"
} }
} }
) )
queries.add( indexes.add(
"svd_wo_whiten", "svd_wo_whiten",
{ {
"gallery_fea_dir": "", "gallery_fea_dir": "",
@ -76,16 +78,17 @@ queries.add(
"feature_names": [], "feature_names": [],
"post_processor": { "dim_processors": {
"name": "PartSVD", "names": ["L2Normalize", "SVD", "L2Normalize"],
"PartSVD": { "SVD": {
"whiten": False, "whiten": False,
"train_fea_dir": "", "train_fea_dir": "",
"proj_dim": 511 "proj_dim": 511,
"l2": True,
} }
}, },
"database_enhance": { "feature_enhancer": {
"name": "Identity" "name": "Identity"
}, },
@ -93,13 +96,13 @@ queries.add(
"name": "KNN" "name": "KNN"
}, },
"re_rank": { "re_ranker": {
"name": "Identity" "name": "Identity"
} }
} }
) )
queries.add( indexes.add(
"svd_whiten", "svd_whiten",
{ {
"gallery_fea_dir": "", "gallery_fea_dir": "",
@ -107,16 +110,17 @@ queries.add(
"feature_names": [], "feature_names": [],
"post_processor": { "dim_processors": {
"name": "PartSVD", "names": ["L2Normalize", "SVD", "L2Normalize"],
"PartSVD": { "SVD": {
"whiten": True, "whiten": True,
"train_fea_dir": "", "train_fea_dir": "",
"proj_dim": 511 "proj_dim": 511,
"l2": True,
} }
}, },
"database_enhance": { "feature_enhancer": {
"name": "Identity" "name": "Identity"
}, },
@ -124,7 +128,7 @@ queries.add(
"name": "KNN" "name": "KNN"
}, },
"re_rank": { "re_ranker": {
"name": "Identity" "name": "Identity"
} }
} }
@ -150,5 +154,5 @@ evaluates.add(
cfg = get_defaults_cfg() cfg = get_defaults_cfg()
queries.check_valid(cfg["query"]) indexes.check_valid(cfg["index"])
evaluates.check_valid(cfg["evaluate"]) evaluates.check_valid(cfg["evaluate"])

14
search/search_rmac_modules/data_process_dict.py → search/search_pwa_modules/pre_process_dict.py
View File

@ -3,9 +3,9 @@
from utils.search_modules import SearchModules from utils.search_modules import SearchModules
from pyretri.config import get_defaults_cfg from pyretri.config import get_defaults_cfg
data_processes = SearchModules() pre_processes = SearchModules()
data_processes.add( pre_processes.add(
"Shorter256Center224", "Shorter256Center224",
{ {
"batch_size": 32, "batch_size": 32,
@ -16,8 +16,8 @@ data_processes.add(
"name": "CollateFn" "name": "CollateFn"
}, },
"transformers": { "transformers": {
"names": ["ResizeShorter", "CenterCrop", "ToTensor", "Normalize"], "names": ["ShorterResize", "CenterCrop", "ToTensor", "Normalize"],
"ResizeShorter": { "ShorterResize": {
"size": 256 "size": 256
}, },
"CenterCrop": { "CenterCrop": {
@ -31,7 +31,7 @@ data_processes.add(
} }
) )
data_processes.add( pre_processes.add(
"Direct224", "Direct224",
{ {
"batch_size": 32, "batch_size": 32,
@ -54,7 +54,7 @@ data_processes.add(
} }
) )
data_processes.add( pre_processes.add(
"PadResize224", "PadResize224",
{ {
"batch_size": 32, "batch_size": 32,
@ -80,4 +80,4 @@ data_processes.add(
cfg = get_defaults_cfg() cfg = get_defaults_cfg()
data_processes.check_valid(cfg["datasets"]) pre_processes.check_valid(cfg["datasets"])

167
search/search_query.py
View File

@ -1,167 +0,0 @@
# -*- coding: utf-8 -*-
import json
import importlib
import os
import argparse
from pyretri.config import get_defaults_cfg
from pyretri.query import build_query_helper
from pyretri.evaluate import build_evaluate_helper
from pyretri.index import feature_loader
vgg_fea = ["pool5_PWA"]
res_fea = ["pool5_PWA"]
task_mapping = {
"oxford_gallery": {
"gallery": "oxford_gallery",
"query": "oxford_query",
"train_fea_dir": "paris"
},
"cub_gallery": {
"gallery": "cub_gallery",
"query": "cub_query",
"train_fea_dir": "cub_gallery"
},
"indoor_gallery": {
"gallery": "indoor_gallery",
"query": "indoor_query",
"train_fea_dir": "indoor_gallery"
},
"caltech101_gallery": {
"gallery": "caltech101_gallery",
"query": "caltech101_query",
"train_fea_dir": "caltech101_gallery"
}
}
def check_exist(now_res, exist_results):
for e_r in exist_results:
totoal_equal = True
for key in now_res:
if now_res[key] != e_r[key]:
totoal_equal = False
break
if totoal_equal:
return True
return False
def get_default_result_dict(dir, task_name, query_name, fea_name):
result_dict = {
"task_name": task_name.split("_")[0],
"dataprocess": dir.split("_")[0],
"model_name": "_".join(dir.split("_")[-2:]),
"feature_map_name": fea_name.split("_")[0],
"fea_process_name": query_name
}
if fea_name == "fc":
result_dict["aggregator_name"] = "none"
else:
result_dict["aggregator_name"] = fea_name.split("_")[1]
return result_dict
def parse_args():
parser = argparse.ArgumentParser(description='A tool box for deep learning-based image retrieval')
parser.add_argument('--fea_dir', '-fd', default=None, type=str, help="path of feature dirs", required=True)
parser.add_argument("--search_modules", "-sm", default=None, type=str, help="name of search module's directory")
parser.add_argument("--save_path", "-sp", default=None, type=str, help="path for saving results")
args = parser.parse_args()
return args
def main():
# init args
args = parse_args()
assert args.fea_dir is not None, 'the feature directory must be provided!'
assert args.search_modules is not None, 'the search modules must be provided!'
assert args.save_path is not None, 'the save path must be provided!'
# init retrieval pipeline settings
cfg = get_defaults_cfg()
# load search space
queries = importlib.import_module("{}.query_dict".format(args.search_modules)).queries
evaluates = importlib.import_module("{}.query_dict".format(args.search_modules)).evaluates
if os.path.exists(args.save_path):
with open(args.save_path, "r") as f:
results = json.load(f)
else:
results = list()
q_cnt = 0
for dir in os.listdir(args.fea_dir):
q_cnt += 1
print("Processing {} / {} queries...".format(q_cnt, len(queries)))
for query_name, query_args in queries.items():
for task_name in task_mapping:
if task_name in dir:
if "vgg" in gallery_fea_dir:
fea_names = vgg_fea
else:
fea_names = res_fea
for fea_name in fea_names:
gallery_fea_dir = os.path.join(args.fea_dir, dir)
query_fea_dir = gallery_fea_dir.replace(task_name, task_mapping[task_name]["query"])
train_fea_dir = gallery_fea_dir.replace(task_name, task_mapping[task_name]["train_fea_dir"])
for post_proc in ["PartPCA", "PartSVD"]:
if post_proc in query_args.post_processors.names:
query_args.post_processors[post_proc].train_fea_dir = train_fea_dir
query.gallery_fea_dir, query.query_fea_dir = gallery_fea_dir, query_fea_dir
query.feature_names = [fea_name]
if task_name == "oxford_base":
evaluate = evaluates["oxford_overall"]
else:
evaluate = evaluates["overall"]
result_dict = get_default_result_dict(dir, task_name, query_name, fea_name)
if check_exist(result_dict, results):
print("[Search Query]: config exists...")
continue
# load retrieval pipeline settings
cfg.query.merge_from_other_cfg(query)
cfg.evaluate.merge_from_other_cfg(evaluate)
# load features
query_fea, query_info, _ = feature_loader.load(cfg.query.query_fea_dir, cfg.query.feature_names)
gallery_fea, gallery_info, _ = feature_loader.load(cfg.query.gallery_fea_dir,
cfg.query.feature_names)
# build helper and index features
query_helper = build_query_helper(cfg.query)
query_result_info, _, _ = query_helper.do_query(query_fea, query_info, gallery_fea)
# build helper and evaluate results
evaluate_helper = build_evaluate_helper(cfg.evaluate)
mAP, recall_at_k = evaluate_helper.do_eval(query_result_info, gallery_info)
# save results
to_save_dict = dict()
for k in recall_at_k:
to_save_dict[str(k)] = recall_at_k[k]
result_dict["mAP"] = float(mAP)
result_dict["recall_at_k"] = to_save_dict
results.append(result_dict)
with open(args.save_path, "w") as f:
json.dump(results, f)
if __name__ == '__main__':
main()

64
search/search_rmac_modules/query_dict.py → search/search_rmac_modules/index_dict.py
View File

@ -3,10 +3,10 @@
from utils.search_modules import SearchModules from utils.search_modules import SearchModules
from pyretri.config import get_defaults_cfg from pyretri.config import get_defaults_cfg
queries = SearchModules() indexes = SearchModules()
evaluates = SearchModules() evaluates = SearchModules()
queries.add( indexes.add(
"pca_wo_whiten", "pca_wo_whiten",
{ {
"gallery_fea_dir": "", "gallery_fea_dir": "",
@ -14,16 +14,17 @@ queries.add(
"feature_names": [], "feature_names": [],
"post_processor": { "dim_processors": {
"name": "PartPCA", "names": ["L2Normalize", "PCA", "L2Normalize"],
"PartPCA": { "PCA": {
"whiten": False, "whiten": False,
"train_fea_dir": "", "train_fea_dir": "",
"proj_dim": 512 "proj_dim": 512,
"l2": True,
} }
}, },
"database_enhance": { "feature_enhancer": {
"name": "Identity" "name": "Identity"
}, },
@ -31,13 +32,13 @@ queries.add(
"name": "KNN" "name": "KNN"
}, },
"re_rank": { "re_ranker": {
"name": "Identity" "name": "Identity"
} }
} }
) )
queries.add( indexes.add(
"pca_whiten", "pca_whiten",
{ {
"gallery_fea_dir": "", "gallery_fea_dir": "",
@ -45,16 +46,17 @@ queries.add(
"feature_names": [], "feature_names": [],
"post_processor": { "dim_processors": {
"name": "PartPCA", "names": ["L2Normalize", "PCA", "L2Normalize"],
"PartPCA": { "PCA": {
"whiten": True, "whiten": True,
"train_fea_dir": "", "train_fea_dir": "",
"proj_dim": 512 "proj_dim": 512,
"l2": True,
} }
}, },
"database_enhance": { "feature_enhancer": {
"name": "Identity" "name": "Identity"
}, },
@ -62,13 +64,13 @@ queries.add(
"name": "KNN" "name": "KNN"
}, },
"re_rank": { "re_ranker": {
"name": "Identity" "name": "Identity"
} }
} }
) )
queries.add( indexes.add(
"svd_wo_whiten", "svd_wo_whiten",
{ {
"gallery_fea_dir": "", "gallery_fea_dir": "",
@ -76,16 +78,17 @@ queries.add(
"feature_names": [], "feature_names": [],
"post_processor": { "dim_processors": {
"name": "PartSVD", "names": ["L2Normalize", "SVD", "L2Normalize"],
"PartSVD": { "SVD": {
"whiten": False, "whiten": False,
"train_fea_dir": "", "train_fea_dir": "",
"proj_dim": 511 "proj_dim": 511,
"l2": True,
} }
}, },
"database_enhance": { "feature_enhancer": {
"name": "Identity" "name": "Identity"
}, },
@ -93,13 +96,13 @@ queries.add(
"name": "KNN" "name": "KNN"
}, },
"re_rank": { "re_ranker": {
"name": "Identity" "name": "Identity"
} }
} }
) )
queries.add( indexes.add(
"svd_whiten", "svd_whiten",
{ {
"gallery_fea_dir": "", "gallery_fea_dir": "",
@ -107,16 +110,17 @@ queries.add(
"feature_names": [], "feature_names": [],
"post_processor": { "dim_processors": {
"name": "PartSVD", "names": ["L2Normalize", "SVD", "L2Normalize"],
"PartSVD": { "SVD": {
"whiten": True, "whiten": True,
"train_fea_dir": "", "train_fea_dir": "",
"proj_dim": 511 "proj_dim": 511,
"l2": True,
} }
}, },
"database_enhance": { "feature_enhancer": {
"name": "Identity" "name": "Identity"
}, },
@ -124,7 +128,7 @@ queries.add(
"name": "KNN" "name": "KNN"
}, },
"re_rank": { "re_ranker": {
"name": "Identity" "name": "Identity"
} }
} }
@ -150,5 +154,5 @@ evaluates.add(
cfg = get_defaults_cfg() cfg = get_defaults_cfg()
queries.check_valid(cfg["query"]) indexes.check_valid(cfg["index"])
evaluates.check_valid(cfg["evaluate"]) evaluates.check_valid(cfg["evaluate"])

14
search/search_pwa_modules/data_process_dict.py → search/search_rmac_modules/pre_process_dict.py
View File

@ -3,9 +3,9 @@
from utils.search_modules import SearchModules from utils.search_modules import SearchModules
from pyretri.config import get_defaults_cfg from pyretri.config import get_defaults_cfg
data_processes = SearchModules() pre_processes = SearchModules()
data_processes.add( pre_processes.add(
"Shorter256Center224", "Shorter256Center224",
{ {
"batch_size": 32, "batch_size": 32,
@ -16,8 +16,8 @@ data_processes.add(
"name": "CollateFn" "name": "CollateFn"
}, },
"transformers": { "transformers": {
"names": ["ResizeShorter", "CenterCrop", "ToTensor", "Normalize"], "names": ["ShorterResize", "CenterCrop", "ToTensor", "Normalize"],
"ResizeShorter": { "ShorterResize": {
"size": 256 "size": 256
}, },
"CenterCrop": { "CenterCrop": {
@ -31,7 +31,7 @@ data_processes.add(
} }
) )
data_processes.add( pre_processes.add(
"Direct224", "Direct224",
{ {
"batch_size": 32, "batch_size": 32,
@ -54,7 +54,7 @@ data_processes.add(
} }
) )
data_processes.add( pre_processes.add(
"PadResize224", "PadResize224",
{ {
"batch_size": 32, "batch_size": 32,
@ -80,4 +80,4 @@ data_processes.add(
cfg = get_defaults_cfg() cfg = get_defaults_cfg()
data_processes.check_valid(cfg["datasets"]) pre_processes.check_valid(cfg["datasets"])

61
search/show_search_results.py
View File

@ -4,9 +4,10 @@ import os
import argparse import argparse
import json import json
import csv
import codecs import codecs
from utils.misc import save_to_csv, filter_by_keywords
def parse_args(): def parse_args():
parser = argparse.ArgumentParser(description='A tool box for deep learning-based image retrieval') parser = argparse.ArgumentParser(description='A tool box for deep learning-based image retrieval')
@ -22,62 +23,30 @@ def show_results(results):
print(results[i]) print(results[i])
def save_to_csv(results, csv_path):
start = []
col_num = 12
if not os.path.exists(csv_path):
start = ["data_process", "model", "feature", "fea_process", "market_mAP", "market_mAP_re", "market_R1",
"market_R1_re", "duke_mAP", "duke_mAP_re", "duke_R1", "duke_R1_re"]
with open(csv_path, 'a+') as f:
csv_write = csv.writer(f)
if len(start) > 0:
csv_write.writerow(start)
for i in range(len(results)):
data_row = [0 for x in range(col_num)]
data_row[0] = results[i]["dataprocess"]
data_row[1] = results[i]["model_name"]
data_row[2] = results[i]["feature_map_name"]
data_row[3] = results[i]["fea_process_name"]
if results[i]["task_name"] == 'market':
data_row[4] = results[i]["mAP"]
data_row[6] = results[i]["recall_at_k"]['1']
elif results[i]["task_name"] == 'duke':
data_row[8] = results[i]["mAP"]
data_row[10] = results[i]["recall_at_k"]['1']
csv_write.writerow(data_row)
def main(): def main():
# init args
args = parse_args() args = parse_args()
assert os.path.exists(args.results_json_path), 'the config file must be existed!' assert os.path.exists(args.results_json_path), 'the config file must be existed!'
with open(args.results_json_path, "r") as f: with open(args.results_json_path, "r") as f:
results = json.load(f) results = json.load(f)
key_words = { # save the search results in a csv format file.
'task_name': ['market'],
'dataprocess': list(),
'model_name': list(),
'feature_map_name': list(),
'aggregator_name': list(),
'fea_process_name': ['no_fea_process', 'l2_normalize', 'pca_whiten', 'pca_wo_whiten'],
}
csv_path = '/home/songrenjie/projects/RetrievalToolBox/test.csv' csv_path = '/home/songrenjie/projects/RetrievalToolBox/test.csv'
save_to_csv(results, csv_path) save_to_csv(results, csv_path)
for key in key_words: # define the keywords to be selected
no_match = [] keywords = {
if len(key_words[key]) == 0: 'data_name': ['market'],
continue 'pre_process_name': list(),
else: 'model_name': list(),
for i in range(len(results)): 'feature_map_name': list(),
if not results[i][key] in key_words[key]: 'aggregator_name': list(),
no_match.append(i) 'post_process_name': ['no_fea_process', 'l2_normalize', 'pca_whiten', 'pca_wo_whiten'],
for num in no_match[::-1]: }
results.pop(num)
# show search results according to the given keywords
results = filter_by_keywords(results, keywords)
show_results(results) show_results(results)

69
search/utils/misc.py
View File

@ -1,9 +1,10 @@
# -*- coding: utf-8 -*- # -*- coding: utf-8 -*-
import os
from typing import Dict, List from typing import Dict, List
import csv
def check_result_exist(now_res: Dict, exist_results: List) -> bool:
def check_exist(now_res: Dict, exist_results: List) -> bool:
""" """
Check if the config exists. Check if the config exists.
@ -45,14 +46,14 @@ def get_dir(root_path: str, dir: str, dataset: Dict) -> (str, str, str):
return gallery_fea_dir, query_fea_dir, train_fea_dir return gallery_fea_dir, query_fea_dir, train_fea_dir
def get_default_result_dict(dir, data_name, query_name, fea_name) -> Dict: def get_default_result_dict(dir: str, data_name: str, index_name: str, fea_name: str) -> Dict:
""" """
Get the default result dict based on the experimental factors. Get the default result dict based on the experimental factors.
Args: Args:
dir (str): the path of one single extracted feature directory. dir (str): the path of one single extracted feature directory.
data_name (str): the name of the dataset. data_name (str): the name of the dataset.
query_name (str): the name of query process. index_name (str): the name of query process.
fea_name (str): the name of the features to be loaded. fea_name (str): the name of the features to be loaded.
Returns: Returns:
@ -60,16 +61,70 @@ def get_default_result_dict(dir, data_name, query_name, fea_name) -> Dict:
""" """
result_dict = { result_dict = {
"data_name": data_name.split("_")[0], "data_name": data_name.split("_")[0],
"dataprocess": dir.split("_")[0], "pre_process_name": dir.split("_")[2],
"model_name": "_".join(dir.split("_")[-2:]), "model_name": "_".join(dir.split("_")[-2:]),
"feature_map_name": fea_name.split("_")[0], "feature_map_name": fea_name.split("_")[0],
"fea_process_name": query_name "post_process_name": index_name
} }
if fea_name == "fc": if len(fea_name.split("_")) == 1:
result_dict["aggregator_name"] = "none" result_dict["aggregator_name"] = "none"
else: else:
result_dict["aggregator_name"] = fea_name.split("_")[1] result_dict["aggregator_name"] = fea_name.split("_")[1]
return result_dict return result_dict
def save_to_csv(results: List[Dict], csv_path: str) -> None:
"""
Save the search results in a csv format file.
Args:
results (List): a list of retrieval results.
csv_path (str): the path for saving the csv file.
"""
start = ["data", "pre_process", "model", "feature_map", "aggregator", "post_process"]
for i in range(len(start)):
results = sorted(results, key=lambda result: result[start[len(start) - i - 1] + "_name"])
start.append('mAP')
start.append('Recall@1')
with open(csv_path, 'w') as f:
csv_write = csv.writer(f)
if len(start) > 0:
csv_write.writerow(start)
for i in range(len(results)):
data_row = [0 for x in range(len(start))]
data_row[0] = results[i]["data_name"]
data_row[1] = results[i]["pre_process_name"]
data_row[2] = results[i]["model_name"]
data_row[3] = results[i]["feature_map_name"]
data_row[4] = results[i]["aggregator_name"]
data_row[5] = results[i]["post_process_name"]
data_row[6] = results[i]["mAP"]
data_row[7] = results[i]["recall_at_k"]['1']
csv_write.writerow(data_row)
def filter_by_keywords(results: List[Dict], keywords: Dict) -> List[Dict]:
"""
Filter the search results according to the given keywords
Args:
results (List): a list of retrieval results.
keywords (Dict): a dict containing keywords to be selected.
Returns:
"""
for key in keywords:
no_match = []
if len(keywords[key]) == 0:
continue
else:
for i in range(len(results)):
if not results[i][key] in keywords[key]:
no_match.append(i)
for num in no_match[::-1]:
results.pop(num)
return results