Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,50 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: mit
|
3 |
+
---
|
4 |
+
# TAC RGB encoder
|
5 |
+
|
6 |
+
<!-- Provide a quick summary of what the model is/does. -->
|
7 |
+
|
8 |
+
This model is used for encoding RGB image into a dense feature.
|
9 |
+
|
10 |
+
**Caution,** the model does not contain the last FC layer.
|
11 |
+
So, the output features are not aligned with depth.
|
12 |
+
|
13 |
+
## Model Details
|
14 |
+
|
15 |
+
### Model Description
|
16 |
+
|
17 |
+
<!-- Provide a longer summary of what this model is. -->
|
18 |
+
|
19 |
+
The model is pre-trained with RGB-D contrastive objectives, named TAC.
|
20 |
+
Different from InfoNCE-based loss fuctions, TAC leverages the similarity between videos frames and estimate a similarity matrix as soft labels.
|
21 |
+
The backbone of this version is ViT-B/32.
|
22 |
+
The pre-training is conducted on a new unified RGB-D database, UniRGBD.
|
23 |
+
The main purpose of this work is depth representation.
|
24 |
+
So, the RGB encoder is just a side model.
|
25 |
+
|
26 |
+
### Model Sources
|
27 |
+
|
28 |
+
<!-- Provide the basic links for the model. -->
|
29 |
+
|
30 |
+
- **Repository:** [TAC](https://github.com/RavenKiller/TAC)
|
31 |
+
- **Paper:** [Learning Depth Representation from RGB-D Videos by Time-Aware Contrastive Pre-training](https://ieeexplore.ieee.org/document/10288539)
|
32 |
+
|
33 |
+
|
34 |
+
## Citation
|
35 |
+
|
36 |
+
<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
|
37 |
+
|
38 |
+
```
|
39 |
+
@ARTICLE{10288539,
|
40 |
+
author={He, Zongtao and Wang, Liuyi and Dang, Ronghao and Li, Shu and Yan, Qingqing and Liu, Chengju and Chen, Qijun},
|
41 |
+
journal={IEEE Transactions on Circuits and Systems for Video Technology},
|
42 |
+
title={Learning Depth Representation from RGB-D Videos by Time-Aware Contrastive Pre-training},
|
43 |
+
year={2023},
|
44 |
+
volume={},
|
45 |
+
number={},
|
46 |
+
pages={1-1},
|
47 |
+
doi={10.1109/TCSVT.2023.3326373}}
|
48 |
+
```
|
49 |
+
|
50 |
+
|