[CodeCamp2023-608] Add Adabins model (#3257)

projects/Adabins/README.md

@@ -0,0 +1,46 @@
+# AdaBins: Depth Estimation Using Adaptive Bins
+
+## Reference
+
+> [AdaBins: Depth Estimation Using Adaptive Bins](https://arxiv.org/abs/2011.14141)
+
+## Introduction
+
+<a href="https://github.com/shariqfarooq123/AdaBins">Official Repo</a>
+
+<a href="https://github.com/open-mmlab/mmsegmentation/tree/dev-1.x/projects/Adabins">Code Snippet</a>
+
+## <img src="https://user-images.githubusercontent.com/34859558/190043857-bfbdaf8b-d2dc-4fff-81c7-e0aac50851f9.png" width="25"/> Abstract
+
+We address the problem of estimating a high quality dense depth map from a single RGB input image. We start out with a baseline encoder-decoder convolutional neural network architecture and pose the question of how the global processing of information can help improve overall depth estimation. To this end, we propose a transformer-based architecture block that divides the depth range into bins whose center value is estimated adaptively per image. The final depth values are estimated as linear combinations of the bin centers. We call our new building block AdaBins. Our results show a decisive improvement over the state-of-the-art on several popular depth datasets across all metrics.We also validate the effectiveness of the proposed block with an ablation study and provide the code and corresponding pre-trained weights of the new state-of-the-art model.
+
+Our main contributions are the following:
+
+- We propose an architecture building block that performs global processing of the scene’s information.We propose to divide the predicted depth range into bins where the bin widths change per image. The final depth estimation is a linear combination of the bin center values.
+- We show a decisive improvement for supervised single image depth estimation across all metrics for the two most popular datasets, NYU and KITTI.
+- We analyze our findings and investigate different modifications on the proposed AdaBins block and study their effect on the accuracy of the depth estimation.
+
+<div align="center">
+<img src="https://github.com/open-mmlab/mmsegmentation/assets/15952744/915bcd5a-9dc2-4602-a6e7-055ff5d4889f"  width = "1000" />
+</div>
+
+## <img src="https://user-images.githubusercontent.com/34859558/190044217-8f6befc2-7f20-473d-b356-148e06265205.png" width="25"/> Performance
+
+### NYU and KITTI
+
+| Model         | Encoder         | Training epoch | Batchsize | Train Resolution | δ1    | δ2    | δ3    | REL   | RMS   | RMS log | params(M) | Links                                                                                                                   |
+| ------------- | --------------- | -------------- | --------- | ---------------- | ----- | ----- | ----- | ----- | ----- | ------- | --------- | ----------------------------------------------------------------------------------------------------------------------- |
+| AdaBins_nyu   | EfficientNet-B5 | 25             | 16        | 416x544          | 0.903 | 0.984 | 0.997 | 0.103 | 0.364 | 0.044   | 78        | [model](https://download.openmmlab.com/mmsegmentation/v0.5/adabins/adabins_efficient_b5_nyu_third-party-f68d6bd3.pth)   |
+| AdaBins_kitti | EfficientNet-B5 | 25             | 16        | 352x764          | 0.964 | 0.995 | 0.999 | 0.058 | 2.360 | 0.088   | 78        | [model](https://download.openmmlab.com/mmsegmentation/v0.5/adabins/adabins_efficient-b5_kitty_third-party-a1aa6f36.pth) |
+
+## Citation
+
+```bibtex
+@article{10.1109/cvpr46437.2021.00400,
+    author = {Bhat, S. A. and Alhashim, I. and Wonka, P.},
+    title = {Adabins: depth estimation using adaptive bins},
+    journal = {2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
+    year = {2021},
+    doi = {10.1109/cvpr46437.2021.00400}
+}
+```

projects/Adabins/backbones/__init__.py

@@ -0,0 +1,4 @@
+# Copyright (c) OpenMMLab. All rights reserved.
+from .adabins_backbone import AdabinsBackbone
+
+__all__ = ['AdabinsBackbone']

projects/Adabins/backbones/adabins_backbone.py

@@ -0,0 +1,141 @@
+import timm
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from mmcv.cnn import ConvModule, build_conv_layer
+from mmengine.model import BaseModule
+
+from mmseg.registry import MODELS
+
+
+class UpSampleBN(nn.Module):
+    """ UpSample module
+    Args:
+        skip_input (int): the input feature
+        output_features (int): the output feature
+        norm_cfg (dict, optional): Config dict for normalization layer.
+            Default: dict(type='BN', requires_grad=True).
+        act_cfg (dict, optional): The activation layer of AAM:
+            Aggregate Attention Module.
+    """
+
+    def __init__(self,
+                 skip_input,
+                 output_features,
+                 norm_cfg=dict(type='BN'),
+                 act_cfg=dict(type='LeakyReLU')):
+        super().__init__()
+
+        self._net = nn.Sequential(
+            ConvModule(
+                in_channels=skip_input,
+                out_channels=output_features,
+                kernel_size=3,
+                stride=1,
+                padding=1,
+                bias=True,
+                norm_cfg=norm_cfg,
+                act_cfg=act_cfg,
+            ),
+            ConvModule(
+                in_channels=output_features,
+                out_channels=output_features,
+                kernel_size=3,
+                stride=1,
+                padding=1,
+                bias=True,
+                norm_cfg=norm_cfg,
+                act_cfg=act_cfg,
+            ))
+
+    def forward(self, x, concat_with):
+        up_x = F.interpolate(
+            x,
+            size=[concat_with.size(2),
+                  concat_with.size(3)],
+            mode='bilinear',
+            align_corners=True)
+        f = torch.cat([up_x, concat_with], dim=1)
+        return self._net(f)
+
+
+class Encoder(nn.Module):
+    """ the efficientnet_b5 model
+    Args:
+        basemodel_name (str): the name of base model
+    """
+
+    def __init__(self, basemodel_name):
+        super().__init__()
+        self.original_model = timm.create_model(
+            basemodel_name, pretrained=True)
+        # Remove last layer
+        self.original_model.global_pool = nn.Identity()
+        self.original_model.classifier = nn.Identity()
+
+    def forward(self, x):
+        features = [x]
+        for k, v in self.original_model._modules.items():
+            if k == 'blocks':
+                for ki, vi in v._modules.items():
+                    features.append(vi(features[-1]))
+            else:
+                features.append(v(features[-1]))
+        return features
+
+
+@MODELS.register_module()
+class AdabinsBackbone(BaseModule):
+    """ the backbone of the adabins
+    Args:
+        basemodel_name (str):the name of base model
+        num_features (int): the middle feature
+        num_classes (int): the classes number
+        bottleneck_features (int): the bottleneck features
+        conv_cfg (dict): Config dict for convolution layer.
+    """
+
+    def __init__(self,
+                 basemodel_name,
+                 num_features=2048,
+                 num_classes=128,
+                 bottleneck_features=2048,
+                 conv_cfg=dict(type='Conv')):
+        super().__init__()
+        self.encoder = Encoder(basemodel_name)
+        features = int(num_features)
+        self.conv2 = build_conv_layer(
+            conv_cfg,
+            bottleneck_features,
+            features,
+            kernel_size=1,
+            stride=1,
+            padding=1)
+        self.up1 = UpSampleBN(
+            skip_input=features // 1 + 112 + 64, output_features=features // 2)
+        self.up2 = UpSampleBN(
+            skip_input=features // 2 + 40 + 24, output_features=features // 4)
+        self.up3 = UpSampleBN(
+            skip_input=features // 4 + 24 + 16, output_features=features // 8)
+        self.up4 = UpSampleBN(
+            skip_input=features // 8 + 16 + 8, output_features=features // 16)
+
+        self.conv3 = build_conv_layer(
+            conv_cfg,
+            features // 16,
+            num_classes,
+            kernel_size=3,
+            stride=1,
+            padding=1)
+
+    def forward(self, x):
+        features = self.encoder(x)
+        x_block0, x_block1, x_block2, x_block3, x_block4 = features[
+            3], features[4], features[5], features[7], features[10]
+        x_d0 = self.conv2(x_block4)
+        x_d1 = self.up1(x_d0, x_block3)
+        x_d2 = self.up2(x_d1, x_block2)
+        x_d3 = self.up3(x_d2, x_block1)
+        x_d4 = self.up4(x_d3, x_block0)
+        out = self.conv3(x_d4)
+        return out

projects/Adabins/configs/_base_/datasets/nyu.py

@@ -0,0 +1,32 @@
+dataset_type = 'NYUDataset'
+data_root = 'data/nyu'
+
+test_pipeline = [
+    dict(dict(type='LoadImageFromFile', to_float32=True)),
+    dict(dict(type='LoadDepthAnnotation', depth_rescale_factor=1e-3)),
+    dict(
+        type='PackSegInputs',
+        meta_keys=('img_path', 'depth_map_path', 'ori_shape', 'img_shape',
+                   'pad_shape', 'scale_factor', 'flip', 'flip_direction',
+                   'category_id'))
+]
+
+val_dataloader = dict(
+    batch_size=1,
+    num_workers=4,
+    persistent_workers=True,
+    sampler=dict(type='DefaultSampler', shuffle=False),
+    dataset=dict(
+        type=dataset_type,
+        data_root=data_root,
+        test_mode=True,
+        data_prefix=dict(
+            img_path='images/test', depth_map_path='annotations/test'),
+        pipeline=test_pipeline))
+test_dataloader = val_dataloader
+
+val_evaluator = dict(
+    type='DepthMetric', max_depth_eval=10.0, crop_type='nyu_crop')
+test_evaluator = val_evaluator
+val_cfg = dict(type='ValLoop')
+test_cfg = dict(type='TestLoop')

projects/Adabins/configs/_base_/default_runtime.py

@@ -0,0 +1,15 @@
+default_scope = 'mmseg'
+env_cfg = dict(
+    cudnn_benchmark=True,
+    mp_cfg=dict(mp_start_method='fork', opencv_num_threads=0),
+    dist_cfg=dict(backend='nccl'),
+)
+vis_backends = [dict(type='LocalVisBackend')]
+visualizer = dict(
+    type='SegLocalVisualizer', vis_backends=vis_backends, name='visualizer')
+log_processor = dict(by_epoch=False)
+log_level = 'INFO'
+load_from = None
+resume = False
+
+tta_model = dict(type='SegTTAModel')

projects/Adabins/configs/_base_/models/Adabins.py

@@ -0,0 +1,35 @@
+# model settings
+norm_cfg = dict(type='SyncBN', requires_grad=True)
+data_preprocessor = dict(
+    type='SegDataPreProcessor',
+    mean=[123.675, 116.28, 103.53],
+    std=[58.395, 57.12, 57.375],
+    bgr_to_rgb=True,
+    pad_val=0,
+    seg_pad_val=255)
+model = dict(
+    type='DepthEstimator',
+    data_preprocessor=data_preprocessor,
+    # pretrained='open-mmlab://resnet50_v1c',
+    backbone=dict(
+        type='AdabinsBackbone',
+        basemodel_name='tf_efficientnet_b5_ap',
+        num_features=2048,
+        num_classes=128,
+        bottleneck_features=2048,
+    ),
+    decode_head=dict(
+        type='AdabinsHead',
+        in_channels=128,
+        n_query_channels=128,
+        patch_size=16,
+        embedding_dim=128,
+        num_heads=4,
+        n_bins=256,
+        min_val=0.001,
+        max_val=10,
+        norm='linear'),
+
+    # model training and testing settings
+    train_cfg=dict(),
+    test_cfg=dict(mode='whole'))

projects/Adabins/configs/adabins/adabins_efficient_b5_4x16_25e_NYU_416x544.py

@@ -0,0 +1,15 @@
+_base_ = [
+    '../_base_/models/Adabins.py', '../_base_/datasets/nyu.py',
+    '../_base_/default_runtime.py'
+]
+custom_imports = dict(
+    imports=['projects.Adabins.backbones', 'projects.Adabins.decode_head'],
+    allow_failed_imports=False)
+crop_size = (416, 544)
+data_preprocessor = dict(size=crop_size)
+norm_cfg = dict(type='SyncBN', requires_grad=True)
+model = dict(
+    data_preprocessor=data_preprocessor,
+    backbone=dict(),
+    decode_head=dict(),
+)

projects/Adabins/configs/adabins/adabins_efficient_b5_4x16_25e_kitti_352x704.py

@@ -0,0 +1,12 @@
+_base_ = ['../_base_/models/Adabins.py']
+custom_imports = dict(
+    imports=['projects.Adabins.backbones', 'projects.Adabins.decode_head'],
+    allow_failed_imports=False)
+crop_size = (352, 704)
+data_preprocessor = dict(size=crop_size)
+norm_cfg = dict(type='SyncBN', requires_grad=True)
+model = dict(
+    data_preprocessor=data_preprocessor,
+    backbone=dict(),
+    decode_head=dict(min_val=0.001, max_val=80),
+)

projects/Adabins/decode_head/__init__.py

@@ -0,0 +1,4 @@
+# Copyright (c) OpenMMLab. All rights reserved.
+from .adabins_head import AdabinsHead
+
+__all__ = ['AdabinsHead']