merge discrim into discrim-final, set as new branch

SimarKareer · Jan 3, 2024 · 92b576a · 92b576a
2 parents b2c2654 + 9aa6761
commit 92b576a
Show file tree

Hide file tree

Showing 30 changed files with 1,504 additions and 262 deletions.
diff --git a/.gitignore b/.gitignore
@@ -221,4 +221,4 @@ tools/exps/mmv1/mmHRDA.sh
 tools/exps/mmv3/mmv3RGB.sh
 tools/exps/sourceEval/sourceEval.sh
 tools/exps/sourceEval/sourceEval2.sh
-tools/exps
+testing/
diff --git a/configs/_base_/datasets/source_BDDSeq.py b/configs/_base_/datasets/source_BDDSeq.py
@@ -0,0 +1,129 @@
+# dataset settings
+FRAME_OFFSET = -2
+dataset_type = 'BDDSeqDataset'
+bdd_data_root = '/coc/flash9/datasets/bdd100k/videoda-subset'
+
+# Backward
+bdd_train_flow_dir= "/coc/flash9/datasets/bdd100k_flow/t_t-1/frame_dist_2/backward/train/images"
+bdd_val_flow_dir = "/coc/flash9/datasets/bdd100k_flow/t_t-1/frame_dist_2/backward/val/images"
+
+img_norm_cfg = dict(
+    mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
+
+crop_size = (720, 720)
+ignore_index = [5, 3, 16, 12, 201, 255] # viper
+# ignore_index = [3, 4, 9, 14, 15, 16, 17, 18, 201, 255] # synthia
+
+
+bdd_train_pipeline = {
+    "im_load_pipeline": [
+        dict(type='LoadImageFromFile'),
+        dict(type='LoadAnnotations'),
+    ],
+    "load_no_ann_pipeline": [
+        dict(type='LoadImageFromFile'),
+    ],
+    "load_flow_pipeline": [
+        dict(type='LoadFlowFromFile'),
+    ],
+    "shared_pipeline": [
+        dict(type='Resize', img_scale=(1280, 720)), # not sure since bdd is 720 x 1280
+        dict(type='RandomCrop', crop_size=crop_size),
+        dict(type='RandomFlip', prob=0.5),
+    ],
+    "im_pipeline": [
+        # dict(type='PhotoMetricDistortion'),
+        dict(type='Normalize', **img_norm_cfg),
+        dict(type='Pad', size=crop_size, pad_val=0, seg_pad_val=255),
+        # dict(type='DefaultFormatBundle'), 
+        dict(type='ImageToTensor', keys=['img']),
+        dict(type='Collect', keys=['img', 'gt_semantic_seg']),
+    ],
+    "flow_pipeline": [
+        dict(type='Pad', size=crop_size, pad_val=0, seg_pad_val=255),
+        dict(type='DefaultFormatBundle'),
+        dict(type='Collect', keys=['img', 'gt_semantic_seg']),
+    ]
+}
+
+test_pipeline = {
+    "im_load_pipeline": [
+        dict(type='LoadImageFromFile'),
+        dict(type='LoadAnnotations'),
+    ],
+    "load_no_ann_pipeline": [
+        dict(type='LoadImageFromFile'),
+    ],
+    "load_flow_pipeline": [
+        dict(type='LoadFlowFromFile'),
+    ],
+    "shared_pipeline": [
+        dict(type='Resize', keep_ratio=True, img_scale=(1280, 720)),
+        # dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
+        dict(type='RandomFlip', prob=0.0),
+    ],
+    "im_pipeline": [
+        # dict(type='PhotoMetricDistortion'),
+        dict(type='Normalize', **img_norm_cfg),
+        # dict(type='Pad', size=crop_size, pad_val=0, seg_pad_val=255),
+        # dict(type='DefaultFormatBundle'),
+        dict(type='ImageToTensor', keys=['img']),
+        dict(type='Collect', keys=['img', 'gt_semantic_seg'])#, meta_keys=[]),
+    ],
+    "flow_pipeline": [
+        # dict(type='Pad', size=crop_size, pad_val=0, seg_pad_val=255),
+        dict(type='DefaultFormatBundle'), #I don't know what this is
+        dict(type='Collect', keys=['img', 'gt_semantic_seg'])#, meta_keys=[]),
+    ]
+}
+
+
+data = dict(
+    train=dict(
+        type='UDADataset',
+        source=dict(
+            type='BDDSeqDataset',
+            data_root=bdd_data_root,
+            img_dir='train_orig_10k/images',
+            ann_dir='train_orig_10k/labels',
+            split='splits/valid_imgs_train.txt',
+            pipeline=bdd_train_pipeline,
+            frame_offset=FRAME_OFFSET,
+            flow_dir=bdd_train_flow_dir,
+            ignore_index=ignore_index,
+        ),
+        target=dict(
+            type='BDDSeqDataset',
+            data_root=bdd_data_root,
+            img_dir='train_orig_10k/images',
+            ann_dir='train_orig_10k/labels',
+            split='splits/valid_imgs_train.txt',
+            pipeline=bdd_train_pipeline,
+            frame_offset=FRAME_OFFSET,
+            flow_dir=bdd_train_flow_dir,
+            ignore_index=ignore_index,
+        )
+    ),
+    val=dict(
+        type='BDDSeqDataset',
+        data_root=bdd_data_root,
+        img_dir='val_orig_10k/images',
+        ann_dir='val_orig_10k/labels',
+        split='splits/valid_imgs_val.txt',
+        pipeline=test_pipeline,
+        frame_offset=FRAME_OFFSET,
+        flow_dir=bdd_val_flow_dir,
+        ignore_index=ignore_index
+    ),
+    test=dict(
+        type='BDDSeqDataset',
+        data_root=bdd_data_root,
+        img_dir='val_orig_10k/images',
+        ann_dir='val_orig_10k/labels',
+        split='splits/valid_imgs_val.txt',
+        pipeline=test_pipeline,
+        frame_offset=FRAME_OFFSET,
+        flow_dir=bdd_val_flow_dir,
+        ignore_index=ignore_index
+    )
+)
diff --git a/configs/_base_/datasets/uda_synthiaSeq_BDDSeq.py b/configs/_base_/datasets/uda_synthiaSeq_BDDSeq.py
@@ -0,0 +1,161 @@
+# dataset settings
+FRAME_OFFSET = -2
+dataset_type = 'SynthiaSeqDataset'
+synthia_data_root = '/srv/share4/datasets/SynthiaSeq/SYNTHIA-SEQS-04-DAWN'
+bdd_data_root = '/coc/flash9/datasets/bdd100k/videoda-subset'
+
+synthia_train_flow_dir = '/srv/share4/datasets/SynthiaSeq_Flow/frame_dist_1/forward/train/RGB/Stereo_Left/Omni_F'
+
+# Backward
+bdd_train_flow_dir= "/coc/flash9/datasets/bdd100k_flow/t_t-1/frame_dist_2/backward/train/images"
+bdd_val_flow_dir = "/coc/flash9/datasets/bdd100k_flow/t_t-1/frame_dist_2/backward/val_orig_10k/images"
+
+img_norm_cfg = dict(
+    mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375], to_rgb=True)
+
+crop_size = (720, 720)
+ignore_index = [3, 4, 9, 14, 15, 16, 17, 18, 201, 255] 
+
+synthia_train_pipeline = {
+    "im_load_pipeline": [
+        dict(type='LoadImageFromFile'),
+        dict(type='LoadAnnotations'),
+    ],
+    "load_no_ann_pipeline": [
+        dict(type='LoadImageFromFile'),
+    ],
+    "load_flow_pipeline": [
+        dict(type='LoadFlowFromFile'),
+    ],
+    "shared_pipeline": [
+        dict(type='Resize', img_scale=(2560, 1520)),
+        dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
+        dict(type='RandomFlip', prob=0.5),
+    ],
+    "im_pipeline": [
+        # dict(type='PhotoMetricDistortion'),
+        dict(type='Normalize', **img_norm_cfg),
+        dict(type='Pad', size=crop_size, pad_val=0, seg_pad_val=255),
+        # dict(type='DefaultFormatBundle'), #I'm not sure why I had to comment it for im, but not for flow.
+        dict(type='ImageToTensor', keys=['img']),
+        dict(type='Collect', keys=['img', 'gt_semantic_seg']),
+    ],
+    "flow_pipeline": [
+        dict(type='Pad', size=crop_size, pad_val=0, seg_pad_val=255),
+        dict(type='DefaultFormatBundle'), #I don't know what this is
+        dict(type='Collect', keys=['img', 'gt_semantic_seg']),
+    ]
+}
+
+
+bdd_train_pipeline = {
+    "im_load_pipeline": [
+        dict(type='LoadImageFromFile'),
+        dict(type='LoadAnnotations'),
+    ],
+    "load_no_ann_pipeline": [
+        dict(type='LoadImageFromFile'),
+    ],
+    "load_flow_pipeline": [
+        dict(type='LoadFlowFromFile'),
+    ],
+    "shared_pipeline": [
+        dict(type='Resize', img_scale=(1280, 720)), # not sure since bdd is 720 x 1280
+        dict(type='RandomCrop', crop_size=crop_size),
+        dict(type='RandomFlip', prob=0.5),
+    ],
+    "im_pipeline": [
+        # dict(type='PhotoMetricDistortion'),
+        dict(type='Normalize', **img_norm_cfg),
+        dict(type='Pad', size=crop_size, pad_val=0, seg_pad_val=255),
+        # dict(type='DefaultFormatBundle'), 
+        dict(type='ImageToTensor', keys=['img']),
+        dict(type='Collect', keys=['img', 'gt_semantic_seg']),
+    ],
+    "flow_pipeline": [
+        dict(type='Pad', size=crop_size, pad_val=0, seg_pad_val=255),
+        dict(type='DefaultFormatBundle'),
+        dict(type='Collect', keys=['img', 'gt_semantic_seg']),
+    ]
+}
+
+test_pipeline = {
+    "im_load_pipeline": [
+        dict(type='LoadImageFromFile'),
+        dict(type='LoadAnnotations'),
+    ],
+    "load_no_ann_pipeline": [
+        dict(type='LoadImageFromFile'),
+    ],
+    "load_flow_pipeline": [
+        dict(type='LoadFlowFromFile'),
+    ],
+    "shared_pipeline": [
+        dict(type='Resize', keep_ratio=True, img_scale=(1280, 720)),
+        # dict(type='RandomCrop', crop_size=crop_size, cat_max_ratio=0.75),
+        dict(type='RandomFlip', prob=0.0),
+    ],
+    "im_pipeline": [
+        # dict(type='PhotoMetricDistortion'),
+        dict(type='Normalize', **img_norm_cfg),
+        # dict(type='Pad', size=crop_size, pad_val=0, seg_pad_val=255),
+        # dict(type='DefaultFormatBundle'),
+        dict(type='ImageToTensor', keys=['img']),
+        dict(type='Collect', keys=['img', 'gt_semantic_seg'])#, meta_keys=[]),
+    ],
+    "flow_pipeline": [
+        # dict(type='Pad', size=crop_size, pad_val=0, seg_pad_val=255),
+        dict(type='DefaultFormatBundle'), #I don't know what this is
+        dict(type='Collect', keys=['img', 'gt_semantic_seg'])#, meta_keys=[]),
+    ]
+}
+
+
+data = dict(
+    train=dict(
+        type='UDADataset',
+        source=dict(
+            type='SynthiaSeqDataset',
+            data_root=synthia_data_root,
+            img_dir='RGB/Stereo_Left/Omni_F',
+            ann_dir='GT/LABELS/Stereo_Left/Omni_F',
+            split='splits/flow/forward/train.txt',
+            pipeline=synthia_train_pipeline,
+            frame_offset=1,
+            flow_dir=synthia_train_flow_dir, 
+        ),
+        target=dict(
+            type='BDDSeqDataset',
+            data_root=bdd_data_root,
+            img_dir='train/images',
+            ann_dir='train/labels',
+            split='splits/valid_imgs_train.txt',
+            pipeline=bdd_train_pipeline,
+            frame_offset=FRAME_OFFSET,
+            flow_dir=bdd_train_flow_dir,
+            ignore_index=ignore_index,
+        )
+    ),
+    val=dict(
+        type='BDDSeqDataset',
+        data_root=bdd_data_root,
+        img_dir='val_orig_10k/images',
+        ann_dir='val_orig_10k/labels',
+        split='splits/valid_imgs_val.txt',
+        pipeline=test_pipeline,
+        frame_offset=FRAME_OFFSET,
+        flow_dir=bdd_val_flow_dir,
+        ignore_index=ignore_index
+    ),
+    test=dict(
+        type='BDDSeqDataset',
+        data_root=bdd_data_root,
+        img_dir='val_orig_10k/images',
+        ann_dir='val_orig_10k/labels',
+        split='splits/valid_imgs_val.txt',
+        pipeline=test_pipeline,
+        frame_offset=FRAME_OFFSET,
+        flow_dir=bdd_val_flow_dir,
+        ignore_index=ignore_index
+    )
+)