fix model deploy problems

demo/README.md

@@ -2,9 +2,18 @@
 
 We provide a command line tool to run a simple demo of builtin models.
 
-You can run this command to get cosine similarites between different images 
+You can run this command to get rank visualization results by cosine similarites between different images.
 
-```bash
-cd demo/
-sh run_demo.sh
+```shell script
+python3 demo/visualize_result.py --config-file logs/dukemtmc/mgn_R50-ibn/config.yaml \
+--parallel --vis-label --dataset-name 'DukeMTMC' --output logs/mgn_duke_vis \
+--opts MODEL.WEIGHTS logs/dukemtmc/mgn_R50-ibn/model_final.pth
+```
+
+You can also run this command to extract image features.
+
+```shell script
+python3 demo/demo.py --config-file logs/dukemtmc/sbs_R50/config.yaml \
+--parallel --input tools/deploy/test_data/*.jpg --output sbs_R50_feat \
+--opts MODEL.WEIGHTS logs/dukemtmc/sbs_R50/model_final.pth
 ```

demo/demo.py

@@ -14,7 +14,7 @@ import numpy as np
 import tqdm
 from torch.backends import cudnn
 
-sys.path.append('..')
+sys.path.append('.')
 
 from fastreid.config import get_cfg
 from fastreid.utils.logger import setup_logger
@@ -86,4 +86,4 @@ if __name__ == '__main__':
             img = cv2.imread(path)
             feat = demo.run_on_image(img)
             feat = feat.numpy()
-            np.save(os.path.join(args.output, os.path.basename(path) + '.npy'), feat)
+            np.save(os.path.join(args.output, os.path.basename(path).split('.')[0] + '.npy'), feat)

demo/run_demo.sh

@@ -1,3 +0,0 @@
-python demo/visualize_result.py --config-file logs/dukemtmc/mgn_R50-ibn/config.yaml \
---parallel --vis-label --dataset-name 'DukeMTMC' --output logs/mgn_duke_vis \
---opts MODEL.WEIGHTS logs/dukemtmc/mgn_R50-ibn/model_final.pth

tools/deploy/README.md

@@ -18,14 +18,16 @@ This is a tiny example for converting fastreid-baseline in `meta_arch` to Caffe
 1. Run `caffe_export.py` to get the converted Caffe model,
 
     ```bash
-    python caffe_export.py --config-file root-path/market1501/bagtricks_R50/config.yml --name baseline_R50 --output outputs/caffe_model --opts MODEL.WEIGHTS root-path/logs/market1501/bagtricks_R50/model_final.pth
+    python tools/deploy/caffe_export.py --config-file configs/market1501/bagtricks_R50/config.yml --name baseline_R50 --output caffe_R50_model --opts MODEL.WEIGHTS logs/market1501/bagtricks_R50/model_final.pth
     ```
 
-    then you can check the Caffe model and prototxt in `outputs/caffe_model`.
+    then you can check the Caffe model and prototxt in `./caffe_R50_model`.
 
-2. Change `prototxt` following next two steps:
+2. Change `prototxt` following next three steps:
 
-   1) Modify `avg_pooling` in `baseline_R50.prototxt`
+   1) Modify `MaxPooling` in `baseline_R50.prototxt` and delete `ceil_mode: false`.
+   
+   2) Add `avg_pooling` in `baseline_R50.prototxt`
 
         ```prototxt
         layer {

tools/deploy/caffe_export.py

@@ -10,7 +10,7 @@ import sys
 
 import torch
 
-sys.path.append('../../')
+sys.path.append('.')
 
 import pytorch_to_caffe
 from fastreid.config import get_cfg
@@ -20,8 +20,8 @@ from fastreid.utils.checkpoint import Checkpointer
 from fastreid.utils.logger import setup_logger
 
 # import some modules added in project like this below
-# sys.path.append('../projects/FastCls')
-# from fastcls import *
+# sys.path.append("projects/PartialReID")
+# from partialreid import *
 
 setup_logger(name='fastreid')
 logger = logging.getLogger("fastreid.caffe_export")
@@ -29,7 +29,6 @@ logger = logging.getLogger("fastreid.caffe_export")
 
 def setup_cfg(args):
     cfg = get_cfg()
-    # add_cls_config(cfg)
     cfg.merge_from_file(args.config_file)
     cfg.merge_from_list(args.opts)
     cfg.freeze()
@@ -69,8 +68,7 @@ if __name__ == '__main__':
 
     cfg.defrost()
     cfg.MODEL.BACKBONE.PRETRAIN = False
-    if cfg.MODEL.HEADS.POOL_LAYER == 'FastGlobalAvgPool':
-        cfg.MODEL.HEADS.POOL_LAYER = 'GlobalAvgPool'
+    cfg.MODEL.HEADS.POOL_LAYER = "Identity"
     cfg.MODEL.BACKBONE.WITH_NL = False
 
     model = build_model(cfg)

tools/deploy/caffe_inference.py

@@ -91,5 +91,5 @@ if __name__ == "__main__":
             net.blobs["blob1"].data[...] = image
             feat = net.forward()["output"]
             feat = normalize(feat[..., 0, 0], axis=1)
-            np.save(os.path.join(args.output, path.replace('.jpg', '.npy').split('/')[-1]), feat)
+            np.save(os.path.join(args.output, os.path.basename(path).split('.')[0] + '.npy'), feat)
 

tools/deploy/onnx_export.py

@@ -4,6 +4,7 @@
 @contact: sherlockliao01@gmail.com
 """
 
+import logging
 import os
 import argparse
 import io
@@ -15,7 +16,7 @@ import torch
 from onnxsim import simplify
 from torch.onnx import OperatorExportTypes
 
-sys.path.append('../../')
+sys.path.append('.')
 
 from fastreid.config import get_cfg
 from fastreid.modeling.meta_arch import build_model
@@ -24,10 +25,12 @@ from fastreid.utils.checkpoint import Checkpointer
 from fastreid.utils.logger import setup_logger
 
 # import some modules added in project like this below
-# sys.path.append('../../projects/FastDistill')
+# sys.path.append("projects/FastDistill")
 # from fastdistill import *
 
-logger = setup_logger(name='onnx_export')
+
+setup_logger(name="fastreid")
+logger = logging.getLogger("fastreid.onnx_export")
 
 
 def setup_cfg(args):
@@ -149,7 +152,7 @@ if __name__ == '__main__':
     model.eval()
     logger.info(model)
 
-    inputs = torch.randn(args.batch_size, 3, cfg.INPUT.SIZE_TEST[0], cfg.INPUT.SIZE_TEST[1])
+    inputs = torch.randn(args.batch_size, 3, cfg.INPUT.SIZE_TEST[0], cfg.INPUT.SIZE_TEST[1]).to(model.device)
     onnx_model = export_onnx_model(model, inputs)
 
     model_simp, check = simplify(onnx_model)

tools/deploy/trt_export.py

@@ -12,12 +12,11 @@ import tensorrt as trt
 
 from trt_calibrator import FeatEntropyCalibrator
 
-sys.path.append('../../')
+sys.path.append('.')
 
-from fastreid.utils.logger import setup_logger
-from fastreid.utils.file_io import PathManager
+from fastreid.utils.logger import setup_logger, PathManager
 
-logger = setup_logger(name='trt_export')
+logger = setup_logger(name="trt_export")
 
 
 def get_parser():

tools/deploy/trt_inference.py

@@ -197,4 +197,4 @@ if __name__ == "__main__":
             # the model expects RGB inputs
             cvt_img = img[:, :, ::-1]
             feat = trt.inference_on_images([cvt_img])
-            np.save(os.path.join(args.output, os.path.basename(img_path) + '.npy'), feat[0])
+            np.save(os.path.join(args.output, os.path.basename(img_path).split('.')[0] + '.npy'), feat)