Toward Practical Monocular Indoor Depth Estimation

Cho-Ying Wu, Jialiang Wang, Michael Hall, Ulrich Neumann, Shuochen Su

[arXiv] [CVF open access] [project site: data, supplementary]

Updates

[June 2023]: Revise the instruction for training codes and train on your own dataset.

[June 2023]: Fix bugs in sample training code.

[June 2023]: Fix bugs in visualization and saving.

Introduction

As this project includes data contribution, please refer to the project page for data download instructions, including SimSIN, UniSIN, and VA, as well as UniSIN leaderboard participation.

Advantage

Results

DistDepth

Our DistDepth is a highly robust monocular depth estimation approach for generic indoor scenes.

• Trained with stereo sequences without their groundtruth depth
• Structured and metric-accurate
• Run in an interactive rate with Laptop GPU
• Sim-to-real: trained on simulation and becomes transferrable to real scenes

Single Image Inference Demo

We test on Ubuntu 20.04 LTS with an laptop NVIDIA 2080 GPU.

Install packages

1. Use conda

   conda create --name distdepth python=3.8 conda activate distdepth

2. Install pre-requisite common packages. Go to https://pytorch.org/get-started/locally/ and install pytorch that is compatible to your computer. We test on pytorch v1.9.0 and cudatoolkit-11.1. (The codes should work under other v1.0+ versions)

3. Install other dependencies: opencv-python and matplotlib, imageio, Pillow, augly, tensorboardX

   pip install opencv-python, matplotlib, imageio, Pillow, augly, tensorboardX

Download pretrained models

4. Download pretrained models [here] (ResNet152, 246MB, illustation for averagely good in-the-wild indoor scenes).

5. Unzip the model under the root directory. 'ckpts' containing the pretrained models is then created.

6. Run

   python demo.py

7. Results will be stored under results/

Pointcloud Generation

Some Sample data are provided in data/sample_pc.

python visualize_pc.py

This will generate pointcloud in '.ply' format by image and depth map inputs for 'data/sample_pc/0000.jpg'. ply file is saved under 'data/sample_pc' folder. Use meshlab to visualize the pointcloud.

Data

Download SimSIN [here]. For UniSIN and VA, please download at the [project site].

To generate stereo data with depth using Habitat, we provide a snippet here. Install Habitat first.

python visualize_pc.py

Training with PoseNet and DepthNet

For a simple taste of training, download a smaller replica set [here] and create and put under './SimSIN-simple'.

The folder structure should be

  .
  ├── SimSIN-simple
        ├── replica
        ├── replica_train.txt

Download weights

mkdir weights

wget -O weights/dpt_hybrid_nyu-2ce69ec7.pt https://github.com/intel-isl/DPT/releases/download/1_0/dpt_hybrid_nyu-2ce69ec7.pt 

Training command

The below command trains networks by using stereo and current frame (PoseNet is not used)

python execute.py --exe train --model_name distdepth-distilled --frame_ids 0 --log_dir='./tmp' --data_path SimSIN-simple --dataset SimSIN  --batch_size 12 --width 256 --height 256 --max_depth 10.0  --num_epochs 10 --scheduler_step_size 8 --learning_rate 0.0001 --use_stereo  --thre 0.95 --num_layers 152 --log_frequency 25

The below command trains networks by using current and past/future one frame

python execute.py --exe train --model_name distdepth-distilled --frame_ids 0 -1 1 --log_dir='./tmp' --data_path SimSIN-simple --dataset SimSIN  --batch_size 12 --width 256 --height 256 --max_depth 10.0  --num_epochs 10 --scheduler_step_size 8 --learning_rate 0.0001 --thre 0.95 --num_layers 152 --log_frequency 25

The below command trains networks by using current and past/future one frame and stereo

python execute.py --exe train --model_name distdepth-distilled --frame_ids 0 -1 1 --log_dir='./tmp' --data_path SimSIN-simple --dataset SimSIN  --batch_size 12 --width 256 --height 256 --max_depth 10.0  --num_epochs 10 --scheduler_step_size 8 --learning_rate 0.0001 --thre 0.95 --num_layers 152 --log_frequency 25 --use_stereo

The memory requires about 20 min on a RTX 3090 GPU.

Changing different expert network: See execute_func.py L59. Switch to different version of DPTDepthModel. The default now used DPT finetuned on NYUv2

If you would like to use more frames, you'll need to leave more buffer frames in the data list file. See below notes for details.

Notes for training on your own dataset:

1. Create your dataloader. You can find SimSIN sample (containing both temporal and stereo) under dataset/ , and then add your dataloader in execute_func.py L111.

2. In execute_func.py L130-141, add your data list file. See format in Replica sample data. Specifically each line contains <file_path> <temporal_step> <left_or_right_for_stereo>

3. Use the before commands to train on your data. Note that your data need to have stereo if you specify --use_stereo. If you sepcify frame_id -1, 1, you'll need to leave one buffer frame at the top and end to avoid reading from None. For example, replica sample data contain 0-49 time steps, but in the data list file, only 1-48 are in file

Evaluation

SimSIN trained models, evaluation on VA

Name	Arch	Expert	MAE	AbsRel	RMSE	acc@ 1.25	acc@ 1.25^2	acc@ 1.25^3	Download
DistDepth	ResNet152	DPT Large	0.252	0.175	0.371	75.1	93.9	98.4	model
DistDepth	ResNet152	DPT Legacy	0.270	0.186	0.386	73.2	93.2	97.9	model
DistDepth-Multi	ResNet101	DPT Legacy	0.243	0.169	0.362	77.1	93.7	97.9	model

Download VA (8G) first. Extract under the root folder.

  .
  ├── VA
        ├── camera_0
           ├── 00000000.png 
               ......
        ├── camera_1
           ├── 00000000.png 
               ......
        ├── gt_depth_rectify
           ├── cam0_frame0000.depth.pfm 
               ......
        ├── VA_left_all.txt

Run bash eval.sh The performances will be saved under the root folder.

To visualize the predicted depth maps in a minibatch (adjust batch_size for different numbers):

python execute.py --exe eval_save --log_dir='./tmp' --data_path VA --dataset VA  --batch_size 10 --load_weights_folder <path to weights> --models_to_load encoder depth  --width 256 --height 256 --max_depth 10 --frame_ids 0 --num_layers 152

If missing 'weights/dpt_hybrid_nyu-2ce69ec7.pt' message pops up, download the model from DPT and put it under 'weights'.

To visualize the predicted depth maps for all testing data on the list:

python execute.py --exe eval_save_all --log_dir='./tmp' --data_path VA --dataset VA  --batch_size 1 --load_weights_folder <path to weights> --models_to_load encoder depth  --width 256 --height 256 --max_depth 10 --frame_ids 0 --num_layers 152

Only batch_size = 1 is valid under this mode.

Evaluation on NYUv2

Prepare NYUv2 data.

  .
  ├── NYUv2
        ├── img_val
           ├── 00001.png
           ......
        ├── depth_val
           ├── 00001.npy
           ......
           ......
        ├── NYUv2.txt

Name	Arch	Expert	MAE	AbsRel	RMSE	acc@ 1.25	acc@ 1.25^2	acc@ 1.25^3	Download
DistDepth-finetuned	ResNet152	DPT on NYUv2	0.308	0.113	0.444	87.3	97.3	99.3	model
DistDepth-SimSIN	ResNet152	DPT	0.411	0.163	0.563	78.0	93.6	98.1	model

Change train_filenames (dummy) and val_filenames in execute_func.py to NYUv2. Then,

python execute.py --exe eval_measure --log_dir='./tmp' --data_path NYUv2 --dataset NYUv2  --batch_size 1 --load_weights_folder <path to weights> --models_to_load encoder depth  --width 256 --height 256 --max_depth 12 --frame_ids 0 --num_layers 152

Depth-aware AR effects

Virtual object insertion:

Dragging objects along a trajectory:

Citation

@inproceedings{wu2022toward,
title={Toward Practical Monocular Indoor Depth Estimation},
author={Wu, Cho-Ying and Wang, Jialiang and Hall, Michael and Neumann, Ulrich and Su, Shuochen},
booktitle={CVPR},
year={2022}
}

License

DistDepth is CC-BY-NC licensed, as found in the LICENSE file.