Upload: example config with wandbvisbackend

Author: maengjemo

custom_configs/DELIVER/hinton-deliver_cmnext_rcnn_lr0.01_ep50.py

@@ -0,0 +1,228 @@
+# CMNeXt Detection with RCNN detector
+# custom_configs/DELIVER/deliver_cmnext_rcnn.py
+import os
+_base_ = [
+    './deliver_dataset.py'  # Inherit dataset config
+]
+
+data_root= '/SSDb/jemo_maeng/dset/DELIVER'
+
+# Model settings
+model = dict(
+    type='FasterRCNN',
+    data_preprocessor=_base_.data_preprocessor,  # This comes from _base_
+    backbone=dict(
+        type='CMNextBackbone',
+        backbone='CMNeXt-B2',
+        modals=['rgb', 'depth', 'event', 'lidar'],
+        out_indices=(0, 1, 2, 3),
+        frozen_stages=-1,
+        pretrained='/SSDb/jemo_maeng/src/Project/Drone24/detection/drone-mmdetection-jm/pretrained_weights/segformer/mit_b2.pth'
+    ),
+    neck=dict(
+        type='FPN',  # MMDetection 표준 FPN 사용
+        in_channels=[64, 128, 320, 512],
+        out_channels=256,
+        num_outs=5
+    ),
+    rpn_head=dict(
+        type='RPNHead',
+        in_channels=256,
+        feat_channels=256,
+        anchor_generator=dict(
+            type='AnchorGenerator',  # 32 16 8 4 
+            # scales=[8],
+            scales=[2, 4, 8, 16],
+            ratios=[0.5, 1.0, 2.0],
+            strides=[4, 8, 16, 32, 64]
+        ),
+        bbox_coder=dict(
+            type='DeltaXYWHBBoxCoder',
+            target_means=[.0, .0, .0, .0],
+            target_stds=[1.0, 1.0, 1.0, 1.0]
+        ),
+        loss_cls=dict(
+            type='CrossEntropyLoss', use_sigmoid=True, loss_weight=1.0
+        ),
+        loss_bbox=dict(type='L1Loss', loss_weight=1.0)
+    ),
+    roi_head=dict(
+        type='StandardRoIHead',
+        bbox_roi_extractor=dict(
+            type='SingleRoIExtractor',
+            roi_layer=dict(type='RoIAlign', output_size=7, sampling_ratio=0),
+            out_channels=256,
+            featmap_strides=[4, 8, 16, 32]
+        ),
+        bbox_head=dict(
+            type='Shared2FCBBoxHead',
+            in_channels=256,
+            fc_out_channels=1024,
+            roi_feat_size=7,
+            num_classes=2,  # Vehicle, Human
+            bbox_coder=dict(
+                type='DeltaXYWHBBoxCoder',
+                target_means=[0., 0., 0., 0.],
+                target_stds=[0.1, 0.1, 0.2, 0.2]
+            ),
+            reg_class_agnostic=False,
+            loss_cls=dict(
+                type='CrossEntropyLoss', use_sigmoid=False, loss_weight=1.0
+            ),
+            loss_bbox=dict(type='L1Loss', loss_weight=1.0)
+        )
+    ),
+    # Training config
+    train_cfg=dict(
+        rpn=dict(
+            assigner=dict(
+                type='MaxIoUAssigner',
+                pos_iou_thr=0.7,
+                neg_iou_thr=0.3,
+                min_pos_iou=0.3,
+                match_low_quality=True,
+                ignore_iof_thr=-1
+            ),
+            sampler=dict(
+                type='RandomSampler',
+                num=256,
+                pos_fraction=0.5,
+                neg_pos_ub=-1,
+                add_gt_as_proposals=False
+            ),
+            allowed_border=-1,
+            pos_weight=-1,
+            debug=False
+        ),
+        rpn_proposal=dict(
+            nms_pre=2000,
+            max_per_img=1000,
+            nms=dict(type='nms', iou_threshold=0.7),
+            min_bbox_size=0
+        ),
+        rcnn=dict(
+            assigner=dict(
+                type='MaxIoUAssigner',
+                pos_iou_thr=0.5,        # 🔥 0.5 → 0.3
+                neg_iou_thr=0.5,        # 🔥 0.5 → 0.1  
+                min_pos_iou=0.5,        # 🔥 0.5 → 0.1
+                match_low_quality=False, # 🔥 False → True
+                ignore_iof_thr=-1
+            ),
+            sampler=dict(
+                type='RandomSampler',
+                num=512,
+                pos_fraction=0.25,
+                neg_pos_ub=-1,
+                add_gt_as_proposals=True
+            ),
+            pos_weight=-1,
+            debug=False
+        )
+    ),
+    # Testing config
+    test_cfg=dict(
+        rpn=dict(
+            nms_pre=1000,
+            max_per_img=1000,
+            nms=dict(type='nms', iou_threshold=0.7),
+            min_bbox_size=0
+        ),
+        rcnn=dict(
+            score_thr=0.05,
+            nms=dict(type='nms', iou_threshold=0.5),
+            max_per_img=100
+        )
+    )
+)
+
+train_dataloader = dict(
+    batch_size=8,
+    num_workers=2,
+    persistent_workers=True,
+    sampler=dict(type='DefaultSampler', shuffle=True),
+    dataset=dict(
+        data_root=data_root,
+        ann_file='coco_train_xywh.json',
+    ),
+)
+
+val_dataloader = dict(
+    batch_size=1,
+    num_workers=2,
+    persistent_workers=True,
+    drop_last=False,
+    sampler=dict(type='DefaultSampler', shuffle=False),
+    dataset=dict(
+        data_root=data_root,
+        ann_file='coco_val_xywh.json',
+    ),
+)
+
+test_dataloader = val_dataloader
+
+# Evaluation settings  
+val_evaluator = dict(
+    type='CocoMetric',
+    ann_file=os.path.join(data_root, 'coco_val_xywh.json'),  # Fixed: consistent with dataset
+    metric='bbox',
+    format_only=False
+)
+train_cfg = dict(
+    type='EpochBasedTrainLoop', 
+    max_epochs=50, 
+    val_interval=5)
+
+
+optim_wrapper = dict(
+    type='OptimWrapper',
+    optimizer=dict(type='SGD', lr=0.01, momentum=0.9, weight_decay=0.0001),
+    clip_grad=dict(max_norm=5, norm_type=2),
+    accumulative_counts=4
+)
+
+param_scheduler = [
+    dict(
+        type='LinearLR',
+        start_factor=0.001,
+        by_epoch=False,
+        begin=0,
+        end=500),  # warmup
+    dict(
+        type='CosineAnnealingLR',
+        T_max=100,  # cosine annealing
+        by_epoch=True,
+        begin=10,
+        end=50,
+        eta_min=1e-6)
+]
+
+vis_backends = [
+    dict(type='LocalVisBackend'),
+    dict(
+        type='WandbVisBackend',
+        init_kwargs=dict(
+            project='DELIVER',
+            name='hinton-deliver_cmnext_rcnn_lr0.01_ep50',
+            tags=['CMNeXt', 'RCNN', 'full-finetune', 'epoch-50'],
+            notes='CMNeXt RCNN with epoch 50 cosinelr',
+            save_code=True
+        ),
+        define_metric_cfg=dict(
+            loss_cls=dict(summary='min'),
+            loss_bbox=dict(summary='min'),
+            loss_iou=dict(summary='min'),
+            loss=dict(summary='min'),
+            bbox_mAP=dict(summary='max'),
+            bbox_mAP_50=dict(summary='max'),
+            bbox_mAP_75=dict(summary='max')
+        )
+    )
+]
+
+
+# Experiment name for logging
+experiment_name = 'deliver_cmnext_b2_faster_rcnn_2x_cosinelr0.01_ep50'
+
+# Override work_dir if needed
+work_dir = f'./work_dirs/{experiment_name}'