Rename imgs_per_gpu to samples_per_gpu (open-mmlab#2404)

* rename api

* Refactor (APIs): rename imgs_per_gpu to samples_per_gpu

* Clean files

* Refactor (infos): rename self.infos to self.data_infos

Author: ZwwWayne

.gitignore

@@ -117,3 +117,4 @@ work_dirs/
 # Pytorch
 *.pth
 *.py~
+*.sh~

configs/_base_/datasets/cityscapes_detection.py

@@ -29,7 +29,7 @@
         ])
 ]
 data = dict(
-    imgs_per_gpu=1,
+    samples_per_gpu=1,
     workers_per_gpu=2,
     train=dict(
         type='RepeatDataset',

configs/_base_/datasets/coco_detection.py

@@ -28,7 +28,7 @@
         ])
 ]
 data = dict(
-    imgs_per_gpu=2,
+    samples_per_gpu=2,
     workers_per_gpu=2,
     train=dict(
         type=dataset_type,

configs/_base_/datasets/voc0712.py

@@ -29,7 +29,7 @@
         ])
 ]
 data = dict(
-    imgs_per_gpu=2,
+    samples_per_gpu=2,
     workers_per_gpu=2,
     train=dict(
         type='RepeatDataset',

configs/_base_/datasets/wider_face.py

@@ -40,7 +40,7 @@
         ])
 ]
 data = dict(
-    imgs_per_gpu=60,
+    samples_per_gpu=60,
     workers_per_gpu=2,
     train=dict(
         type='RepeatDataset',

configs/fast_rcnn/fast_rcnn_r50_fpn_1x_coco.py

@@ -39,7 +39,7 @@
         ])
 ]
 data = dict(
-    imgs_per_gpu=2,
+    samples_per_gpu=2,
     workers_per_gpu=2,
     train=dict(
         proposal_file=data_root + 'proposals/rpn_r50_fpn_1x_train2017.pkl',

configs/fcos/fcos_r101_caffe_fpn_gn-head_mstrain_640-800_4x4_2x_coco.py

@@ -33,7 +33,7 @@
         ])
 ]
 data = dict(
-    imgs_per_gpu=4,
+    samples_per_gpu=4,
     workers_per_gpu=4,
     train=dict(pipeline=train_pipeline),
     val=dict(pipeline=test_pipeline),

configs/fcos/fcos_r50_caffe_fpn_gn-head_4x4_1x_coco.py

@@ -85,7 +85,7 @@
         ])
 ]
 data = dict(
-    imgs_per_gpu=4,
+    samples_per_gpu=4,
     workers_per_gpu=4,
     train=dict(pipeline=train_pipeline),
     val=dict(pipeline=test_pipeline),

configs/fcos/fcos_x101_64x4d_fpn_gn-head_mstrain_640-800_4x2_2x_coco.py

@@ -44,7 +44,7 @@
         ])
 ]
 data = dict(
-    imgs_per_gpu=2,
+    samples_per_gpu=2,
     workers_per_gpu=2,
     train=dict(pipeline=train_pipeline),
     val=dict(pipeline=test_pipeline),

configs/foveabox/fovea_r50_fpn_4x4_1x_coco.py

@@ -47,6 +47,6 @@
     score_thr=0.05,
     nms=dict(type='nms', iou_thr=0.5),
     max_per_img=100)
-data = dict(imgs_per_gpu=4, workers_per_gpu=4)
+data = dict(samples_per_gpu=4, workers_per_gpu=4)
 # optimizer
 optimizer = dict(type='SGD', lr=0.01, momentum=0.9, weight_decay=0.0001)

configs/htc/htc_x101_32x4d_fpn_16x1_20e_coco.py

@@ -12,7 +12,7 @@
         norm_cfg=dict(type='BN', requires_grad=True),
         norm_eval=True,
         style='pytorch'))
-data = dict(imgs_per_gpu=1, workers_per_gpu=1)
+data = dict(samples_per_gpu=1, workers_per_gpu=1)
 # learning policy
 lr_config = dict(step=[16, 19])
 total_epochs = 20

configs/htc/htc_x101_64x4d_fpn_16x1_20e_coco.py

@@ -12,7 +12,7 @@
         norm_cfg=dict(type='BN', requires_grad=True),
         norm_eval=True,
         style='pytorch'))
-data = dict(imgs_per_gpu=1, workers_per_gpu=1)
+data = dict(samples_per_gpu=1, workers_per_gpu=1)
 # learning policy
 lr_config = dict(step=[16, 19])
 total_epochs = 20

configs/htc/htc_x101_64x4d_fpn_dconv_c3-c5_mstrain_400_1400_16x1_20e_coco.py

@@ -36,7 +36,7 @@
         keys=['img', 'gt_bboxes', 'gt_labels', 'gt_masks', 'gt_semantic_seg']),
 ]
 data = dict(
-    imgs_per_gpu=1, workers_per_gpu=1, train=dict(pipeline=train_pipeline))
+    samples_per_gpu=1, workers_per_gpu=1, train=dict(pipeline=train_pipeline))
 # learning policy
 lr_config = dict(step=[16, 19])
 total_epochs = 20

configs/nas_fpn/retinanet_r50_fpn_crop640_50e_coco.py

@@ -56,7 +56,7 @@
         ])
 ]
 data = dict(
-    imgs_per_gpu=8,
+    samples_per_gpu=8,
     workers_per_gpu=4,
     train=dict(pipeline=train_pipeline),
     val=dict(pipeline=test_pipeline),

configs/nas_fpn/retinanet_r50_nasfpn_crop640_50e_coco.py

@@ -55,7 +55,7 @@
         ])
 ]
 data = dict(
-    imgs_per_gpu=8,
+    samples_per_gpu=8,
     workers_per_gpu=4,
     train=dict(pipeline=train_pipeline),
     val=dict(pipeline=test_pipeline),

configs/pascal_voc/ssd300_voc0712.py

@@ -45,7 +45,7 @@
         ])
 ]
 data = dict(
-    imgs_per_gpu=8,
+    samples_per_gpu=8,
     workers_per_gpu=3,
     train=dict(
         type='RepeatDataset', times=10, dataset=dict(pipeline=train_pipeline)),

configs/ssd/ssd300_coco.py

@@ -44,7 +44,7 @@
         ])
 ]
 data = dict(
-    imgs_per_gpu=8,
+    samples_per_gpu=8,
     workers_per_gpu=3,
     train=dict(
         _delete_=True,

configs/ssd/ssd512_coco.py

@@ -50,7 +50,7 @@
         ])
 ]
 data = dict(
-    imgs_per_gpu=8,
+    samples_per_gpu=8,
     workers_per_gpu=3,
     train=dict(
         _delete_=True,

mmdet/apis/test.py

@@ -59,7 +59,9 @@ def multi_gpu_test(model, data_loader, tmpdir=None, gpu_collect=False):
         results.append(result)
 
         if rank == 0:
-            batch_size = data['img'][0].size(0)
+            batch_size = (
+                len(data['img_meta']._data)
+                if 'img_meta' in data else data['img'][0].size(0))
             for _ in range(batch_size * world_size):
                 prog_bar.update()
 

mmdet/apis/train.py

@@ -123,7 +123,7 @@ def _dist_train(model,
     data_loaders = [
         build_dataloader(
             ds,
-            cfg.data.imgs_per_gpu,
+            cfg.data.samples_per_gpu,
             cfg.data.workers_per_gpu,
             dist=True,
             seed=cfg.seed) for ds in dataset
@@ -167,7 +167,7 @@ def _dist_train(model,
         val_dataset = build_dataset(cfg.data.val, dict(test_mode=True))
         val_dataloader = build_dataloader(
             val_dataset,
-            imgs_per_gpu=1,
+            samples_per_gpu=1,
             workers_per_gpu=cfg.data.workers_per_gpu,
             dist=True,
             shuffle=False)
@@ -193,7 +193,7 @@ def _non_dist_train(model,
     data_loaders = [
         build_dataloader(
             ds,
-            cfg.data.imgs_per_gpu,
+            cfg.data.samples_per_gpu,
             cfg.data.workers_per_gpu,
             cfg.gpus,
             dist=False,
@@ -228,7 +228,7 @@ def _non_dist_train(model,
         val_dataset = build_dataset(cfg.data.val, dict(test_mode=True))
         val_dataloader = build_dataloader(
             val_dataset,
-            imgs_per_gpu=1,
+            samples_per_gpu=1,
             workers_per_gpu=cfg.data.workers_per_gpu,
             dist=False,
             shuffle=False)

mmdet/datasets/cityscapes.py

@@ -25,7 +25,7 @@ def _filter_imgs(self, min_size=32):
         """Filter images too small or without ground truths."""
         valid_inds = []
         ids_with_ann = set(_['image_id'] for _ in self.coco.anns.values())
-        for i, img_info in enumerate(self.img_infos):
+        for i, img_info in enumerate(self.data_infos):
             img_id = img_info['id']
             ann_ids = self.coco.getAnnIds(imgIds=[img_id])
             ann_info = self.coco.loadAnns(ann_ids)
@@ -113,7 +113,7 @@ def results2txt(self, results, outfile_prefix):
         prog_bar = mmcv.ProgressBar(len(self))
         for idx in range(len(self)):
             result = results[idx]
-            filename = self.img_infos[idx]['filename']
+            filename = self.data_infos[idx]['filename']
             basename = osp.splitext(osp.basename(filename))[0]
             pred_txt = osp.join(outfile_prefix, basename + '_pred.txt')
 

mmdet/datasets/coco.py

@@ -36,24 +36,24 @@ def load_annotations(self, ann_file):
         self.cat_ids = self.coco.getCatIds()
         self.cat2label = {cat_id: i for i, cat_id in enumerate(self.cat_ids)}
         self.img_ids = self.coco.getImgIds()
-        img_infos = []
+        data_infos = []
         for i in self.img_ids:
             info = self.coco.loadImgs([i])[0]
             info['filename'] = info['file_name']
-            img_infos.append(info)
-        return img_infos
+            data_infos.append(info)
+        return data_infos
 
     def get_ann_info(self, idx):
-        img_id = self.img_infos[idx]['id']
+        img_id = self.data_infos[idx]['id']
         ann_ids = self.coco.getAnnIds(imgIds=[img_id])
         ann_info = self.coco.loadAnns(ann_ids)
-        return self._parse_ann_info(self.img_infos[idx], ann_info)
+        return self._parse_ann_info(self.data_infos[idx], ann_info)
 
     def _filter_imgs(self, min_size=32):
         """Filter images too small or without ground truths."""
         valid_inds = []
         ids_with_ann = set(_['image_id'] for _ in self.coco.anns.values())
-        for i, img_info in enumerate(self.img_infos):
+        for i, img_info in enumerate(self.data_infos):
             if self.filter_empty_gt and self.img_ids[i] not in ids_with_ann:
                 continue
             if min(img_info['width'], img_info['height']) >= min_size:

mmdet/datasets/custom.py

@@ -64,15 +64,15 @@ def __init__(self,
                 self.proposal_file = osp.join(self.data_root,
                                               self.proposal_file)
         # load annotations (and proposals)
-        self.img_infos = self.load_annotations(self.ann_file)
+        self.data_infos = self.load_annotations(self.ann_file)
         if self.proposal_file is not None:
             self.proposals = self.load_proposals(self.proposal_file)
         else:
             self.proposals = None
         # filter images too small
         if not test_mode:
             valid_inds = self._filter_imgs()
-            self.img_infos = [self.img_infos[i] for i in valid_inds]
+            self.data_infos = [self.data_infos[i] for i in valid_inds]
             if self.proposals is not None:
                 self.proposals = [self.proposals[i] for i in valid_inds]
         # set group flag for the sampler
@@ -82,7 +82,7 @@ def __init__(self,
         self.pipeline = Compose(pipeline)
 
     def __len__(self):
-        return len(self.img_infos)
+        return len(self.data_infos)
 
     def load_annotations(self, ann_file):
         return mmcv.load(ann_file)
@@ -91,7 +91,7 @@ def load_proposals(self, proposal_file):
         return mmcv.load(proposal_file)
 
     def get_ann_info(self, idx):
-        return self.img_infos[idx]['ann']
+        return self.data_infos[idx]['ann']
 
     def pre_pipeline(self, results):
         results['img_prefix'] = self.img_prefix
@@ -104,7 +104,7 @@ def pre_pipeline(self, results):
     def _filter_imgs(self, min_size=32):
         """Filter images too small."""
         valid_inds = []
-        for i, img_info in enumerate(self.img_infos):
+        for i, img_info in enumerate(self.data_infos):
             if min(img_info['width'], img_info['height']) >= min_size:
                 valid_inds.append(i)
         return valid_inds
@@ -117,7 +117,7 @@ def _set_group_flag(self):
         """
         self.flag = np.zeros(len(self), dtype=np.uint8)
         for i in range(len(self)):
-            img_info = self.img_infos[i]
+            img_info = self.data_infos[i]
             if img_info['width'] / img_info['height'] > 1:
                 self.flag[i] = 1
 
@@ -136,7 +136,7 @@ def __getitem__(self, idx):
             return data
 
     def prepare_train_img(self, idx):
-        img_info = self.img_infos[idx]
+        img_info = self.data_infos[idx]
         ann_info = self.get_ann_info(idx)
         results = dict(img_info=img_info, ann_info=ann_info)
         if self.proposals is not None:
@@ -145,7 +145,7 @@ def prepare_train_img(self, idx):
         return self.pipeline(results)
 
     def prepare_test_img(self, idx):
-        img_info = self.img_infos[idx]
+        img_info = self.data_infos[idx]
         results = dict(img_info=img_info)
         if self.proposals is not None:
             results['proposals'] = self.proposals[idx]

mmdet/datasets/loader/build_loader.py

@@ -19,7 +19,7 @@
 
 
 def build_dataloader(dataset,
-                     imgs_per_gpu,
+                     samples_per_gpu,
                      workers_per_gpu,
                      num_gpus=1,
                      dist=True,
@@ -33,8 +33,8 @@ def build_dataloader(dataset,
 
     Args:
         dataset (Dataset): A PyTorch dataset.
-        imgs_per_gpu (int): Number of images on each GPU, i.e., batch size of
-            each GPU.
+        samples_per_gpu (int): Number of training samples on each GPU, i.e.,
+            batch size of each GPU.
         workers_per_gpu (int): How many subprocesses to use for data loading
             for each GPU.
         num_gpus (int): Number of GPUs. Only used in non-distributed training.
@@ -51,16 +51,16 @@ def build_dataloader(dataset,
         # DistributedGroupSampler will definitely shuffle the data to satisfy
         # that images on each GPU are in the same group
         if shuffle:
-            sampler = DistributedGroupSampler(dataset, imgs_per_gpu,
+            sampler = DistributedGroupSampler(dataset, samples_per_gpu,
                                               world_size, rank)
         else:
             sampler = DistributedSampler(
                 dataset, world_size, rank, shuffle=False)
-        batch_size = imgs_per_gpu
+        batch_size = samples_per_gpu
         num_workers = workers_per_gpu
     else:
-        sampler = GroupSampler(dataset, imgs_per_gpu) if shuffle else None
-        batch_size = num_gpus * imgs_per_gpu
+        sampler = GroupSampler(dataset, samples_per_gpu) if shuffle else None
+        batch_size = num_gpus * samples_per_gpu
         num_workers = num_gpus * workers_per_gpu
 
     init_fn = partial(
@@ -72,7 +72,7 @@ def build_dataloader(dataset,
         batch_size=batch_size,
         sampler=sampler,
         num_workers=num_workers,
-        collate_fn=partial(collate, samples_per_gpu=imgs_per_gpu),
+        collate_fn=partial(collate, samples_per_gpu=samples_per_gpu),
         pin_memory=False,
         worker_init_fn=init_fn,
         **kwargs)

mmdet/datasets/wider_face.py

@@ -20,7 +20,7 @@ def __init__(self, **kwargs):
         super(WIDERFaceDataset, self).__init__(**kwargs)
 
     def load_annotations(self, ann_file):
-        img_infos = []
+        data_infos = []
         img_ids = mmcv.list_from_file(ann_file)
         for img_id in img_ids:
             filename = '{}.jpg'.format(img_id)
@@ -32,11 +32,11 @@ def load_annotations(self, ann_file):
             width = int(size.find('width').text)
             height = int(size.find('height').text)
             folder = root.find('folder').text
-            img_infos.append(
+            data_infos.append(
                 dict(
                     id=img_id,
                     filename=osp.join(folder, filename),
                     width=width,
                     height=height))
 
-        return img_infos
+        return data_infos