added a support of SODA-D dataset

configs/yolov7/yolov7_tiny_syncbn_fast_4x16b-100e_sodad.py

@@ -0,0 +1,190 @@
+_base_ = './yolov7_tiny_syncbn_fast_8x16b-300e_coco.py'
+
+# ========================Frequently modified parameters======================
+# -----data related-----
+data_root = './data/SODA-D/'
+train_ann_file = 'Annotations/train.json'
+train_data_prefix = 'Images'  # Prefix of train image path
+# Path of val annotation file
+val_ann_file = 'Annotations/val.json'
+val_data_prefix = 'Images'  # Prefix of val image path
+test_ann_file = 'Annotations/test.json'
+metainfo = dict(
+    classes=('people', 'rider', 'bicycle', 'motor', 'vehicle', 'traffic-sign',
+             'traffic-light', 'traffic-camera', 'warning-cone'),
+    palette=None)
+
+num_classes = 9  # Number of classes for classification
+# Batch size of a single GPU during training
+train_batch_size_per_gpu = 16
+# Worker to pre-fetch data for each single GPU during training
+train_num_workers = 4
+
+# -----model related-----
+# Basic size of multi-scale prior box
+anchors = [
+    [(12, 16), (19, 36), (40, 28)],  # P3/8
+    [(36, 75), (76, 55), (72, 146)],  # P4/16
+    [(142, 110), (192, 243), (459, 401)]  # P5/32
+]
+# -----train val related-----
+# Base learning rate for optim_wrapper. Corresponding to 8xb16=128 bs
+# base_lr = base_lr / num_gpus
+base_lr = 0.01 / 8
+max_epochs = 100  # Maximum training epochs
+
+# ========================Possible modified parameters========================
+# -----data related-----
+img_scale = (1216, 1216)  # width, height
+# Dataset type, this will be used to define the dataset
+dataset_type = 'YOLOv5CocoDataset'
+# Batch size of a single GPU during validation
+val_batch_size_per_gpu = 1
+# Worker to pre-fetch data for each single GPU during validation
+val_num_workers = 2
+
+# Config of batch shapes. Only on val.
+# It means not used if batch_shapes_cfg is None.
+batch_shapes_cfg = dict(
+    batch_size=val_batch_size_per_gpu,
+    img_size=img_scale[0],
+    # The image scale of padding should be divided by pad_size_divisor
+    size_divisor=32)
+
+# -----train val related-----
+loss_cls_weight = 0.3
+loss_bbox_weight = 0.05
+loss_obj_weight = 0.7
+
+# ===============================Unmodified in most cases====================
+model = dict(
+    bbox_head=dict(
+        head_module=dict(num_classes=num_classes),
+        prior_generator=dict(base_sizes=anchors),
+        # scaled based on number of detection layers
+        loss_cls=dict(loss_weight=loss_cls_weight *
+                      (num_classes / 80 * 3 / _base_.num_det_layers)),
+        loss_bbox=dict(
+            loss_weight=loss_bbox_weight * (3 / _base_.num_det_layers),
+            return_iou=True),
+        loss_obj=dict(loss_weight=loss_obj_weight *
+                      ((img_scale[0] / 640)**2 * 3 / _base_.num_det_layers)),
+    ))
+
+
+
+pre_transform = _base_.pre_transform
+max_translate_ratio = _base_.max_translate_ratio
+scaling_ratio_range = _base_.scaling_ratio_range
+randchoice_mosaic_prob = _base_.randchoice_mosaic_prob
+mixup_alpha = _base_.mixup_alpha
+mixup_beta = _base_.mixup_beta
+mixup_prob = _base_.mixup_prob
+
+mosiac4_pipeline = [
+    dict(
+        type='Mosaic',
+        img_scale=img_scale,
+        pad_val=114.0,
+        pre_transform=pre_transform),
+    dict(
+        type='YOLOv5RandomAffine',
+        max_rotate_degree=0.0,
+        max_shear_degree=0.0,
+        max_translate_ratio=max_translate_ratio,  # change
+        scaling_ratio_range=scaling_ratio_range,  # change
+        # img_scale is (width, height)
+        border=(-img_scale[0] // 2, -img_scale[1] // 2),
+        border_val=(114, 114, 114)),
+]
+mosiac9_pipeline = [
+    dict(
+        type='Mosaic9',
+        img_scale=img_scale,
+        pad_val=114.0,
+        pre_transform=pre_transform),
+    dict(
+        type='YOLOv5RandomAffine',
+        max_rotate_degree=0.0,
+        max_shear_degree=0.0,
+        max_translate_ratio=max_translate_ratio,  # change
+        scaling_ratio_range=scaling_ratio_range,  # change
+        border=(-img_scale[0] // 2, -img_scale[1] // 2),
+        border_val=(114, 114, 114)),
+]
+
+randchoice_mosaic_pipeline = dict(
+    type='RandomChoice',
+    transforms=[mosiac4_pipeline, mosiac9_pipeline],
+    prob=randchoice_mosaic_prob)
+
+train_pipeline = [
+    *pre_transform,
+    randchoice_mosaic_pipeline,
+    dict(
+        type='YOLOv5MixUp',
+        alpha=mixup_alpha,
+        beta=mixup_beta,
+        prob=mixup_prob,  # change
+        pre_transform=[*pre_transform, randchoice_mosaic_pipeline]),
+    dict(type='YOLOv5HSVRandomAug'),
+    dict(type='mmdet.RandomFlip', prob=0.5),
+    dict(
+        type='mmdet.PackDetInputs',
+        meta_keys=('img_id', 'img_path', 'ori_shape', 'img_shape', 'flip',
+                   'flip_direction'))
+]
+
+train_dataloader = dict(
+    batch_size=train_batch_size_per_gpu,
+    num_workers=train_num_workers,
+    dataset=dict(
+        type=dataset_type,
+        metainfo=metainfo,
+        data_root=data_root,
+        ann_file=train_ann_file,
+        data_prefix=dict(img=train_data_prefix)))
+
+val_dataloader = dict(
+    batch_size=val_batch_size_per_gpu,
+    num_workers=val_num_workers,
+    dataset=dict(
+        type=dataset_type,
+        metainfo=metainfo,
+        data_root=data_root,
+        data_prefix=dict(img=val_data_prefix),
+        ann_file=val_ann_file,
+        batch_shapes_cfg=batch_shapes_cfg))
+
+test_dataloader = dict(
+    batch_size=val_batch_size_per_gpu,
+    num_workers=val_num_workers,
+    dataset=dict(
+        type=dataset_type,
+        metainfo=metainfo,
+        data_root=data_root,
+        data_prefix=dict(img=val_data_prefix),
+        ann_file=test_ann_file,
+        batch_shapes_cfg=batch_shapes_cfg))
+
+# optimizer
+optim_wrapper = dict(
+    type='OptimWrapper',
+    optimizer=dict(
+        type='SGD',
+        lr=base_lr,
+        momentum=0.9,
+        weight_decay=_base_.weight_decay,
+        nesterov=True,
+        batch_size_per_gpu=train_batch_size_per_gpu),
+    clip_grad=None,
+    constructor='YOLOv7OptimWrapperConstructor')
+
+default_hooks = dict(param_scheduler=dict(max_epochs=max_epochs))
+val_evaluator = dict(type='SodadMetric', classwise=True, ann_file=data_root + val_ann_file)
+test_evaluator = dict(type='SodadMetric', classwise=True, ann_file=data_root + test_ann_file)
+
+train_cfg = dict(
+    max_epochs=max_epochs,
+    dynamic_intervals=[(max_epochs - _base_.num_epoch_stage2,
+                        _base_.val_interval_stage2)])

mmyolo/engine/__init__.py

@@ -1,3 +1,4 @@
 # Copyright (c) OpenMMLab. All rights reserved.
 from .hooks import *  # noqa: F401,F403
 from .optimizers import *  # noqa: F401,F403
+from .evaluation import *  # noqa: F401,F403

mmyolo/engine/evaluation/__init__.py

@@ -0,0 +1,6 @@
+# Copyright (c) OpenMMLab. All rights reserved.
+from .sodad_metric import SodadMetric  # noqa: F401,F403
+
+__all__ = [
+    'SodadMetric'
+]