insailab commited on
Commit
a2142bb
Β·
verified Β·
1 Parent(s): a91b2f2

docs: add direct collection links to organization card

Browse files
Files changed (1) hide show
  1. README.md +53 -0
README.md CHANGED
@@ -119,6 +119,59 @@ hf download InSAI-Lab/360BEV --local-dir ./checkpoints
119
 
120
  ---
121
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
122
  ## πŸ”¬ Recent Selected Publications
123
  - **IEEE T-PAMI 2026:** *OneBEV++: Towards Unifying Bird’s-Eye-View Semantic Mapping with Panoramas*
124
  - **CVPR 2026:** *More than the Sum: Panorama-Language Models for Adverse Omni-Scenes*
 
119
 
120
  ---
121
 
122
+ ### 5. [πŸ“š Collection: Trans4Map (WACV 2023)](https://huggingface.co/collections/InSAI-Lab/trans4map-wacv-2023-6ad1e062eb1718eb4759eb2b)
123
+ > **Incremental Top-Down Semantic Mapping from Egocentric Observations**
124
+
125
+ Transformer-based incremental semantic mapping from egocentric observations to top-down 2D floorplans with bounding sphere attention on Matterport3D.
126
+
127
+ | Resource | Hugging Face Repository | Description |
128
+ | :--- | :--- | :--- |
129
+ | πŸ“¦ **Dataset** | [`InSAI-Lab/Trans4Map`](https://huggingface.co/datasets/InSAI-Lab/Trans4Map) | Matterport3D top-down semantic ground truth package `semmap.tar.gz` (51 MB) |
130
+ | πŸ€– **Model Zoo** | [`InSAI-Lab/Trans4Map`](https://huggingface.co/InSAI-Lab/Trans4Map) | Trained MiT-B2 (119 MB) and MiT-B4 (259 MB) top-down semantic mapping checkpoints |
131
+ | πŸ“„ **Paper** | [arXiv:2207.06205](https://arxiv.org/abs/2207.06205) | IEEE/CVF WACV 2023 Paper |
132
+
133
+ ```bash
134
+ # Quick download via Hugging Face CLI
135
+ hf download InSAI-Lab/Trans4Map semmap.tar.gz --type dataset --local-dir ./data/semmap
136
+ hf download InSAI-Lab/Trans4Map --local-dir ./checkpoints
137
+ ```
138
+
139
+ ---
140
+
141
+ ### 6. [πŸ“š Collection: MatchFormer (ACCV 2022 Oral)](https://huggingface.co/collections/InSAI-Lab/matchformer-accv-2022-6ad1e828aa78e3758323bbc4)
142
+ > **Interleaving Attention in Transformers for Robust Feature Matching**
143
+
144
+ Hierarchical extract-and-match transformer interleaving self- and cross-attention for robust indoor (ScanNet) and outdoor (MegaDepth) local feature matching.
145
+
146
+ | Resource | Hugging Face Repository | Description |
147
+ | :--- | :--- | :--- |
148
+ | πŸ€– **Model Zoo** | [`InSAI-Lab/MatchFormer`](https://huggingface.co/InSAI-Lab/MatchFormer) | Official checkpoints for Indoor (Large SEA, Lite LA) & Outdoor (Large LA, Lite SEA) |
149
+ | πŸ“„ **Paper** | [arXiv:2203.09645](https://arxiv.org/abs/2203.09645) | ACCV 2022 Oral Paper |
150
+
151
+ ```bash
152
+ # Quick download via Hugging Face CLI
153
+ hf download InSAI-Lab/MatchFormer --local-dir ./model/weights
154
+ ```
155
+
156
+ ---
157
+
158
+ ### 7. [πŸ“š Collection: Trans4trans (ICCVW 2021 & IEEE T-ITS 2022)](https://huggingface.co/collections/InSAI-Lab/trans4trans-iccvw-2021-and-t-its-2022-6ad1ed42eaa4c1d9c5832601)
159
+ > **Transparent Object and Semantic Scene Segmentation in Real-World Navigation**
160
+
161
+ Efficient transformer system designed for transparent object segmentation and wearable navigation assistance with Intel RealSense depth sensing and acoustic feedback.
162
+
163
+ | Resource | Hugging Face Repository | Description |
164
+ | :--- | :--- | :--- |
165
+ | πŸ€– **Model Zoo** | [`InSAI-Lab/Trans4trans`](https://huggingface.co/InSAI-Lab/Trans4trans) | Pretrained checkpoints for COCO-Stuff general context and Trans10K-v2 transparent objects |
166
+ | πŸ“„ **Papers** | [arXiv:2107.03172](https://arxiv.org/abs/2107.03172) / [arXiv:2108.09174](https://arxiv.org/abs/2108.09174) | ICCVW 2021 & IEEE T-ITS 2022 |
167
+
168
+ ```bash
169
+ # Quick download via Hugging Face CLI
170
+ hf download InSAI-Lab/Trans4trans --local-dir ./workdirs
171
+ ```
172
+
173
+ ---
174
+
175
  ## πŸ”¬ Recent Selected Publications
176
  - **IEEE T-PAMI 2026:** *OneBEV++: Towards Unifying Bird’s-Eye-View Semantic Mapping with Panoramas*
177
  - **CVPR 2026:** *More than the Sum: Panorama-Language Models for Adverse Omni-Scenes*