Update according to issue #12

Experiment-pretrain_choice-all_tricks-tri_center-market.sh

@@ -0,0 +1,11 @@
+# Experiment all tricks with center loss : 256x128-bs16x4-warmup10-erase0_5-labelsmooth_on-laststride1-bnneck_on-triplet_centerloss0_0005
+# Dataset 1: market1501
+# imagesize: 256x128
+# batchsize: 16x4
+# warmup_step 10
+# random erase prob 0.5
+# labelsmooth: on
+# last stride 1
+# bnneck on
+# with center loss
+python3 tools/train.py --config_file='configs/softmax_triplet_with_center.yml' MODEL.DEVICE_ID "('2')" DATASETS.NAMES "('market1501')" DATASETS.ROOT_DIR "('/home/haoluo/data')" MODEL.PRETRAIN_CHOICE "('self')" MODEL.PRETRAIN_PATH "('/home/haoluo/log/gu/reid_baseline_review/Opensource_test/market1501/Experiment-pretrain_choice_all-tricks-tri_center-256x128-bs16x4-warmup10-erase0_5-labelsmooth_on-laststride1-bnneck_on-triplet_centerloss0_0005/resnet50_model_2.pth')" OUTPUT_DIR "('/home/haoluo/log/gu/reid_baseline_review/Opensource_test/market1501/Experiment-pretrain_choice_all-tricks-tri_center-256x128-bs16x4-warmup10-erase0_5-labelsmooth_on-laststride1-bnneck_on-triplet_centerloss0_0005')"

Test-all_tricks-tri_center-feat_after_bn-cos-duke.sh

@@ -9,4 +9,4 @@
 # bnneck on
 # with center loss
 # without re-ranking
-python3 tools/test.py --config_file='configs/softmax_triplet_with_center.yml' MODEL.DEVICE_ID "('1')" DATASETS.NAMES "('dukemtmc')" DATASETS.ROOT_DIR "('/home/haoluo/data')" TEST.WEIGHT "('/home/haoluo/log/gu/reid_baseline_review/Opensource_test/dukemtmc/Experiment-all-tricks-tri_center-256x128-bs16x4-warmup10-erase0_5-labelsmooth_on-laststride1-bnneck_on-triplet_centerloss0_0005/resnet50_model_120.pth')"
+python3 tools/test.py --config_file='configs/softmax_triplet_with_center.yml' MODEL.DEVICE_ID "('1')" DATASETS.NAMES "('dukemtmc')" DATASETS.ROOT_DIR "('/home/haoluo/data')" MODEL.PRETRAIN_CHOICE "('self')" TEST.WEIGHT "('/home/haoluo/log/gu/reid_baseline_review/Opensource_test/dukemtmc/Experiment-all-tricks-tri_center-256x128-bs16x4-warmup10-erase0_5-labelsmooth_on-laststride1-bnneck_on-triplet_centerloss0_0005/resnet50_model_120.pth')"

Test-all_tricks-tri_center-feat_after_bn-cos-market.sh

@@ -9,4 +9,4 @@
 # bnneck on
 # with center loss
 # without re-ranking
-python3 tools/test.py --config_file='configs/softmax_triplet_with_center.yml' MODEL.DEVICE_ID "('0')" DATASETS.NAMES "('market1501')" DATASETS.ROOT_DIR "('/home/haoluo/data')" TEST.WEIGHT "('/home/haoluo/log/gu/reid_baseline_review/Opensource_test/market1501/Experiment-all-tricks-tri_center-256x128-bs16x4-warmup10-erase0_5-labelsmooth_on-laststride1-bnneck_on-triplet_centerloss0_0005/resnet50_model_120.pth')"
+python3 tools/test.py --config_file='configs/softmax_triplet_with_center.yml' MODEL.DEVICE_ID "('0')" DATASETS.NAMES "('market1501')" DATASETS.ROOT_DIR "('/home/haoluo/data')" MODEL.PRETRAIN_CHOICE "('self')" TEST.WEIGHT "('/home/haoluo/log/gu/reid_baseline_review/Opensource_test/market1501/Experiment-all-tricks-tri_center-256x128-bs16x4-warmup10-erase0_5-labelsmooth_on-laststride1-bnneck_on-triplet_centerloss0_0005/resnet50_model_120.pth')"

Test-all_tricks-without_center-feat_after_bn-cos-duke.sh

@@ -9,4 +9,4 @@
 # bnneck on
 # without center loss
 # without re-ranking
-python3 tools/test.py --config_file='configs/softmax_triplet.yml' MODEL.DEVICE_ID "('1')" DATASETS.NAMES "('dukemtmc')" DATASETS.ROOT_DIR "('/home/haoluo/data')" TEST.WEIGHT "('/home/haoluo/log/gu/reid_baseline_review/Opensource_test/dukemtmc/Experiment-all-tricks-256x128-bs16x4-warmup10-erase0_5-labelsmooth_on-laststride1-bnneck_on/resnet50_model_120.pth')"
+python3 tools/test.py --config_file='configs/softmax_triplet.yml' MODEL.DEVICE_ID "('1')" DATASETS.NAMES "('dukemtmc')" DATASETS.ROOT_DIR "('/home/haoluo/data')" MODEL.PRETRAIN_CHOICE "('self')" TEST.WEIGHT "('/home/haoluo/log/gu/reid_baseline_review/Opensource_test/dukemtmc/Experiment-all-tricks-256x128-bs16x4-warmup10-erase0_5-labelsmooth_on-laststride1-bnneck_on/resnet50_model_120.pth')"

Test-all_tricks-without_center-feat_after_bn-cos-market.sh

@@ -9,4 +9,4 @@
 # bnneck on
 # without center loss
 # without re-ranking
-python3 tools/test.py --config_file='configs/softmax_triplet.yml' MODEL.DEVICE_ID "('0')" DATASETS.NAMES "('market1501')"  DATASETS.ROOT_DIR "('/home/haoluo/data')" TEST.WEIGHT "('/home/haoluo/log/gu/reid_baseline_review/Opensource_test/market1501/Experiment-all-tricks-256x128-bs16x4-warmup10-erase0_5-labelsmooth_on-laststride1-bnneck_on/resnet50_model_120.pth')"
+python3 tools/test.py --config_file='configs/softmax_triplet.yml' MODEL.DEVICE_ID "('0')" DATASETS.NAMES "('market1501')"  DATASETS.ROOT_DIR "('/home/haoluo/data')" MODEL.PRETRAIN_CHOICE "('self')" TEST.WEIGHT "('/home/haoluo/log/gu/reid_baseline_review/Opensource_test/market1501/Experiment-all-tricks-256x128-bs16x4-warmup10-erase0_5-labelsmooth_on-laststride1-bnneck_on/resnet50_model_120.pth')"

Test-reranking-all_tricks-tri_center-feat_after_bn-cos-duke.sh

@@ -9,4 +9,4 @@
 # bnneck on
 # with center loss
 # with re-ranking
-python3 tools/test.py --config_file='configs/softmax_triplet_with_center.yml' MODEL.DEVICE_ID "('1')" DATASETS.NAMES "('dukemtmc')" TEST.RE_RANKING "('yes')"  DATASETS.ROOT_DIR "('/home/haoluo/data')" TEST.WEIGHT "('/home/haoluo/log/gu/reid_baseline_review/Opensource_test/dukemtmc/Experiment-all-tricks-tri_center-256x128-bs16x4-warmup10-erase0_5-labelsmooth_on-laststride1-bnneck_on-triplet_centerloss0_0005/resnet50_model_120.pth')"
+python3 tools/test.py --config_file='configs/softmax_triplet_with_center.yml' MODEL.DEVICE_ID "('1')" DATASETS.NAMES "('dukemtmc')" TEST.RE_RANKING "('yes')"  DATASETS.ROOT_DIR "('/home/haoluo/data')" MODEL.PRETRAIN_CHOICE "('self')" TEST.WEIGHT "('/home/haoluo/log/gu/reid_baseline_review/Opensource_test/dukemtmc/Experiment-all-tricks-tri_center-256x128-bs16x4-warmup10-erase0_5-labelsmooth_on-laststride1-bnneck_on-triplet_centerloss0_0005/resnet50_model_120.pth')"

Test-reranking-all_tricks-tri_center-feat_after_bn-cos-market.sh

@@ -9,4 +9,4 @@
 # bnneck on
 # with center loss
 # with re-ranking
-python3 tools/test.py --config_file='configs/softmax_triplet_with_center.yml' MODEL.DEVICE_ID "('0')" DATASETS.NAMES "('market1501')" DATASETS.ROOT_DIR "('/home/haoluo/data')" TEST.RE_RANKING "('yes')" TEST.WEIGHT "('/home/haoluo/log/gu/reid_baseline_review/Opensource_test/market1501/Experiment-all-tricks-tri_center-256x128-bs16x4-warmup10-erase0_5-labelsmooth_on-laststride1-bnneck_on-triplet_centerloss0_0005/resnet50_model_120.pth')"
+python3 tools/test.py --config_file='configs/softmax_triplet_with_center.yml' MODEL.DEVICE_ID "('0')" DATASETS.NAMES "('market1501')" DATASETS.ROOT_DIR "('/home/haoluo/data')" TEST.RE_RANKING "('yes')" MODEL.PRETRAIN_CHOICE "('self')" TEST.WEIGHT "('/home/haoluo/log/gu/reid_baseline_review/Opensource_test/market1501/Experiment-all-tricks-tri_center-256x128-bs16x4-warmup10-erase0_5-labelsmooth_on-laststride1-bnneck_on-triplet_centerloss0_0005/resnet50_model_120.pth')"

config/defaults.py

@@ -27,6 +27,9 @@ _C.MODEL.NAME = 'resnet50'
 _C.MODEL.LAST_STRIDE = 1
 # Path to pretrained model of backbone
 _C.MODEL.PRETRAIN_PATH = ''
+# Use ImageNet pretrained model to initialize backbone or use self trained model to initialize the whole model
+# Options: 'imagenet' or 'self'
+_C.MODEL.PRETRAIN_CHOICE = 'imagenet'
 # If train with BNNeck, options: 'bnneck' or 'no'
 _C.MODEL.NECK = 'bnneck'
 # If train loss include center loss, options: 'yes' or 'no'. Loss with center loss has different optimizer configuration

configs/baseline.yml

@@ -1,4 +1,5 @@
 MODEL:
+  PRETRAIN_CHOICE: 'imagenet'
   PRETRAIN_PATH: '/home/haoluo/.torch/models/resnet50-19c8e357.pth'
   LAST_STRIDE: 2
   NECK: 'no'

configs/softmax_triplet.yml

@@ -1,4 +1,5 @@
 MODEL:
+  PRETRAIN_CHOICE: 'imagenet'
   PRETRAIN_PATH: '/home/haoluo/.torch/models/resnet50-19c8e357.pth'
   METRIC_LOSS_TYPE: 'triplet'
   IF_LABELSMOOTH: 'on'

configs/softmax_triplet_with_center.yml

@@ -1,4 +1,5 @@
 MODEL:
+  PRETRAIN_CHOICE: 'imagenet'
   PRETRAIN_PATH: '/home/haoluo/.torch/models/resnet50-19c8e357.pth'
   METRIC_LOSS_TYPE: 'triplet_center'
   IF_LABELSMOOTH: 'on'

engine/trainer.py

@@ -130,7 +130,8 @@ def do_train(
         optimizer,
         scheduler,
         loss_fn,
-        num_query
+        num_query,
+        start_epoch
 ):
     log_period = cfg.SOLVER.LOG_PERIOD
     checkpoint_period = cfg.SOLVER.CHECKPOINT_PERIOD
@@ -155,6 +156,10 @@ def do_train(
     RunningAverage(output_transform=lambda x: x[0]).attach(trainer, 'avg_loss')
     RunningAverage(output_transform=lambda x: x[1]).attach(trainer, 'avg_acc')
 
+    @trainer.on(Events.STARTED)
+    def start_training(engine):
+        engine.state.epoch = start_epoch
+
     @trainer.on(Events.EPOCH_STARTED)
     def adjust_learning_rate(engine):
         scheduler.step()
@@ -201,7 +206,8 @@ def do_train_with_center(
         optimizer_center,
         scheduler,
         loss_fn,
-        num_query
+        num_query,
+        start_epoch
 ):
     log_period = cfg.SOLVER.LOG_PERIOD
     checkpoint_period = cfg.SOLVER.CHECKPOINT_PERIOD
@@ -218,7 +224,9 @@ def do_train_with_center(
     timer = Timer(average=True)
 
     trainer.add_event_handler(Events.EPOCH_COMPLETED, checkpointer, {'model': model.state_dict(),
-                                                                     'optimizer': optimizer.state_dict()})
+                                                                     'optimizer': optimizer.state_dict(),
+                                                                     'optimizer_center': optimizer_center.state_dict()})
+
     timer.attach(trainer, start=Events.EPOCH_STARTED, resume=Events.ITERATION_STARTED,
                  pause=Events.ITERATION_COMPLETED, step=Events.ITERATION_COMPLETED)
 
@@ -226,6 +234,10 @@ def do_train_with_center(
     RunningAverage(output_transform=lambda x: x[0]).attach(trainer, 'avg_loss')
     RunningAverage(output_transform=lambda x: x[1]).attach(trainer, 'avg_acc')
 
+    @trainer.on(Events.STARTED)
+    def start_training(engine):
+        engine.state.epoch = start_epoch
+
     @trainer.on(Events.EPOCH_STARTED)
     def adjust_learning_rate(engine):
         scheduler.step()

modeling/__init__.py

@@ -10,5 +10,5 @@ from .baseline import Baseline
 def build_model(cfg, num_classes):
     # if cfg.MODEL.NAME == 'resnet50':
     #     model = Baseline(num_classes, cfg.MODEL.LAST_STRIDE, cfg.MODEL.PRETRAIN_PATH, cfg.MODEL.NECK, cfg.TEST.NECK_FEAT)
-    model = Baseline(num_classes, cfg.MODEL.LAST_STRIDE, cfg.MODEL.PRETRAIN_PATH, cfg.MODEL.NECK, cfg.TEST.NECK_FEAT, cfg.MODEL.NAME)
+    model = Baseline(num_classes, cfg.MODEL.LAST_STRIDE, cfg.MODEL.PRETRAIN_PATH, cfg.MODEL.NECK, cfg.TEST.NECK_FEAT, cfg.MODEL.NAME, cfg.MODEL.PRETRAIN_CHOICE)
     return model

modeling/baseline.py

@@ -37,7 +37,7 @@ def weights_init_classifier(m):
 class Baseline(nn.Module):
     in_planes = 2048
 
-    def __init__(self, num_classes, last_stride, model_path, neck, neck_feat, model_name):
+    def __init__(self, num_classes, last_stride, model_path, neck, neck_feat, model_name, pretrain_choice):
         super(Baseline, self).__init__()
         if model_name == 'resnet18':
             self.base = ResNet(last_stride=last_stride, 
@@ -117,14 +117,14 @@ class Baseline(nn.Module):
                               last_stride=last_stride)
         elif model_name == 'senet154':
             self.base = SENet(block=SEBottleneck, 
-                              layers=[3, 8, 36, 3], 
+                              layers=[3, 8, 36, 3],
                               groups=64, 
                               reduction=16,
                               dropout_p=0.2, 
                               last_stride=last_stride)
 
-            
+        if pretrain_choice == 'imagenet':
-        self.base.load_param(model_path)
+            self.base.load_param(model_path)
         self.gap = nn.AdaptiveAvgPool2d(1)
         # self.gap = nn.AdaptiveMaxPool2d(1)
         self.num_classes = num_classes
@@ -167,6 +167,6 @@ class Baseline(nn.Module):
     def load_param(self, trained_path):
         param_dict = torch.load(trained_path)
         for i in param_dict:
-            if 'classifier' in i:
+            # if 'classifier' in i:
-                continue
+            #     continue
-            self.state_dict()[i].copy_(param_dict[i])
+            self.state_dict()[i].copy_(param_dict[i])

tools/train.py

@@ -7,6 +7,7 @@
 import argparse
 import os
 import sys
+import torch
 
 from torch.backends import cudnn
 
@@ -31,11 +32,28 @@ def train(cfg):
     if cfg.MODEL.IF_WITH_CENTER == 'no':
         print('Train without center loss, the loss type is', cfg.MODEL.METRIC_LOSS_TYPE)
         optimizer = make_optimizer(cfg, model)
-        scheduler = WarmupMultiStepLR(optimizer, cfg.SOLVER.STEPS, cfg.SOLVER.GAMMA, cfg.SOLVER.WARMUP_FACTOR,
+        # scheduler = WarmupMultiStepLR(optimizer, cfg.SOLVER.STEPS, cfg.SOLVER.GAMMA, cfg.SOLVER.WARMUP_FACTOR,
-                                      cfg.SOLVER.WARMUP_ITERS, cfg.SOLVER.WARMUP_METHOD)
+        #                               cfg.SOLVER.WARMUP_ITERS, cfg.SOLVER.WARMUP_METHOD)
 
         loss_func = make_loss(cfg, num_classes)     # modified by gu
 
+        # Add for using self trained model
+        if cfg.MODEL.PRETRAIN_CHOICE == 'self':
+            start_epoch = eval(cfg.MODEL.PRETRAIN_PATH.split('/')[-1].split('.')[0].split('_')[-1])
+            print('Start epoch:', start_epoch)
+            path_to_optimizer = cfg.MODEL.PRETRAIN_PATH.replace('model', 'optimizer')
+            print('Path to the checkpoint of optimizer:', path_to_optimizer)
+            model.load_state_dict(torch.load(cfg.MODEL.PRETRAIN_PATH))
+            optimizer.load_state_dict(torch.load(path_to_optimizer))
+            scheduler = WarmupMultiStepLR(optimizer, cfg.SOLVER.STEPS, cfg.SOLVER.GAMMA, cfg.SOLVER.WARMUP_FACTOR,
+                                          cfg.SOLVER.WARMUP_ITERS, cfg.SOLVER.WARMUP_METHOD, start_epoch)
+        elif cfg.MODEL.PRETRAIN_CHOICE == 'imagenet':
+            start_epoch = 0
+            scheduler = WarmupMultiStepLR(optimizer, cfg.SOLVER.STEPS, cfg.SOLVER.GAMMA, cfg.SOLVER.WARMUP_FACTOR,
+                                          cfg.SOLVER.WARMUP_ITERS, cfg.SOLVER.WARMUP_METHOD)
+        else:
+            print('Only support pretrain_choice for imagenet and self, but got {}'.format(cfg.MODEL.PRETRAIN_CHOICE))
+
         arguments = {}
 
         do_train(
@@ -44,19 +62,40 @@ def train(cfg):
             train_loader,
             val_loader,
             optimizer,
-            scheduler,
+            scheduler,      # modify for using self trained model
             loss_func,
-            num_query
+            num_query,
+            start_epoch     # add for using self trained model
         )
     elif cfg.MODEL.IF_WITH_CENTER == 'yes':
         print('Train with center loss, the loss type is', cfg.MODEL.METRIC_LOSS_TYPE)
         loss_func, center_criterion = make_loss_with_center(cfg, num_classes)  # modified by gu
         optimizer, optimizer_center = make_optimizer_with_center(cfg, model, center_criterion)
-        scheduler = WarmupMultiStepLR(optimizer, cfg.SOLVER.STEPS, cfg.SOLVER.GAMMA, cfg.SOLVER.WARMUP_FACTOR,
+        # scheduler = WarmupMultiStepLR(optimizer, cfg.SOLVER.STEPS, cfg.SOLVER.GAMMA, cfg.SOLVER.WARMUP_FACTOR,
-                                      cfg.SOLVER.WARMUP_ITERS, cfg.SOLVER.WARMUP_METHOD)
+        #                               cfg.SOLVER.WARMUP_ITERS, cfg.SOLVER.WARMUP_METHOD)
 
         arguments = {}
 
+        # Add for using self trained model
+        if cfg.MODEL.PRETRAIN_CHOICE == 'self':
+            start_epoch = eval(cfg.MODEL.PRETRAIN_PATH.split('/')[-1].split('.')[0].split('_')[-1])
+            print('Start epoch:', start_epoch)
+            path_to_optimizer = cfg.MODEL.PRETRAIN_PATH.replace('model', 'optimizer')
+            print('Path to the checkpoint of optimizer:', path_to_optimizer)
+            path_to_optimizer_center = cfg.MODEL.PRETRAIN_PATH.replace('model', 'optimizer_center')
+            print('Path to the checkpoint of optimizer_center:', path_to_optimizer_center)
+            model.load_state_dict(torch.load(cfg.MODEL.PRETRAIN_PATH))
+            optimizer.load_state_dict(torch.load(path_to_optimizer))
+            optimizer_center.load_state_dict(torch.load(path_to_optimizer_center))
+            scheduler = WarmupMultiStepLR(optimizer, cfg.SOLVER.STEPS, cfg.SOLVER.GAMMA, cfg.SOLVER.WARMUP_FACTOR,
+                                          cfg.SOLVER.WARMUP_ITERS, cfg.SOLVER.WARMUP_METHOD, start_epoch)
+        elif cfg.MODEL.PRETRAIN_CHOICE == 'imagenet':
+            start_epoch = 0
+            scheduler = WarmupMultiStepLR(optimizer, cfg.SOLVER.STEPS, cfg.SOLVER.GAMMA, cfg.SOLVER.WARMUP_FACTOR,
+                                          cfg.SOLVER.WARMUP_ITERS, cfg.SOLVER.WARMUP_METHOD)
+        else:
+            print('Only support pretrain_choice for imagenet and self, but got {}'.format(cfg.MODEL.PRETRAIN_CHOICE))
+
         do_train_with_center(
             cfg,
             model,
@@ -65,9 +104,10 @@ def train(cfg):
             val_loader,
             optimizer,
             optimizer_center,
-            scheduler,
+            scheduler,      # modify for using self trained model
             loss_func,
-            num_query
+            num_query,
+            start_epoch     # add for using self trained model
         )
     else:
         print("Unsupported value for cfg.MODEL.IF_WITH_CENTER {}, only support yes or no!\n".format(cfg.MODEL.IF_WITH_CENTER))