Hugging Face's logo

joakimbh
/
ATC

English
Model card Files
xet
Community
joakimbh commited on
Commit
faa5063
·
verified ·
1 Parent(s): 515cc45

Setup model card for ATC model overview

Browse files
Files changed (1) hide show
  1. README.md +103 -3
README.md CHANGED
@@ -1,3 +1,103 @@
1
- ---
2
- license: mit
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: mit
3
+ language:
4
+ - en
5
+ ---
6
+
7
+ # Model Details
8
+ Agglomerative Token Clustering (ATC), a novel hierarichal hard-merging based token reduction method.
9
+ ATC has been demonstrated to outperform all prior state-of-the-art token reduction methods across image classification, image synthesis, and object detection & segementation tasks.
10
+
11
+ We provide DeiT checkpoints (Tiny, Small, and Base) at four reduction rates (0.9, 0.7, 0.5, and 0.25) for four classification datasets: ImageNet-1K, NABirds, COCO 2014, and NUS-WIDE.
12
+ We also provide ViT-Adapter checkpoints (Tiny and Small) for the COCO 2017 object detection and segementation task, for both the ATC and ToMe token reduction methods.
13
+
14
+
15
+
16
+ ### More Resources
17
+ - **Repository:** [https://github.com/JoakimHaurum/ATC](https://github.com/JoakimHaurum/ATC)
18
+ - **Paper:** [TBD](TBD)
19
+ - **Project Page:** [https://vap.aau.dk/atc](https://vap.aau.dk/atc)
20
+ - **Additional Results:** [https://rawalkhirodkar.github.io/sapiens](https://rawalkhirodkar.github.io/sapiens/)
21
+ - **HuggingFace Collection:** [https://huggingface.co/collections/joakimbh/agglomerative-token-clustering-66e94dfb313e85ec97590fe4](https://huggingface.co/collections/joakimbh/agglomerative-token-clustering-66e94dfb313e85ec97590fe4)
22
+
23
+
24
+ ## Model Zoo
25
+ **Note: This repository does not host any checkpoints but contains links to all the model repositories. Each ATC model repository contains checkpoints for the complete, average, and single linkage functions.**
26
+
27
+ Image Classification Models:
28
+ | Model Name | Dataset | Keep Rate | Weights |
29
+ |:-----------|:--------:|:-----------:|:--------:|
30
+ | atc_base-im1k-0.9 | ImageNet-1K | 90% | [link]() |
31
+ | atc_base-im1k-0.7 | ImageNet-1K | 70% | [link]() |
32
+ | atc_base-im1k-0.5 | ImageNet-1K | 50% | [link]() |
33
+ | atc_base-im1k-0.25 | ImageNet-1K | 25% | [link]() |
34
+ | atc_small-im1k-0.9 | ImageNet-1K | 90% | [link]() |
35
+ | atc_small-im1k-0.7 | ImageNet-1K | 70% | [link]() |
36
+ | atc_small-im1k-0.5 | ImageNet-1K | 50% | [link]() |
37
+ | atc_small-im1k-0.25 | ImageNet-1K | 25% | [link]() |
38
+ | atc_tiny-im1k-0.9 | ImageNet-1K | 90% | [link]() |
39
+ | atc_tiny-im1k-0.7 | ImageNet-1K | 70% | [link]() |
40
+ | atc_tiny-im1k-0.5 | ImageNet-1K | 50% | [link]() |
41
+ | atc_tiny-im1k-0.25 | ImageNet-1K | 25% | [link]() |
42
+ <br>
43
+ | atc_base-nab-0.9 | NABirds | 90% | [link]() |
44
+ | atc_base-nab-0.7 | NABirds | 70% | [link]() |
45
+ | atc_base-nab-0.5 | NABirds | 50% | [link]() |
46
+ | atc_base-nab-0.25 | NABirds | 25% | [link]() |
47
+ | atc_small-nab-0.9 | NABirds | 90% | [link]() |
48
+ | atc_small-nab-0.7 | NABirds | 70% | [link]() |
49
+ | atc_small-nab-0.5 | NABirds | 50% | [link]() |
50
+ | atc_small-nab-0.25 | NABirds | 25% | [link]() |
51
+ | atc_tiny-nab-0.9 | NABirds | 90% | [link]() |
52
+ | atc_tiny-nab-0.7 | NABirds | 70% | [link]() |
53
+ | atc_tiny-nab-0.5 | NABirds | 50% | [link]() |
54
+ | atc_tiny-nab-0.25 | NABirds | 25% | [link]() |
55
+ <br>
56
+ | atc_base-coco-0.9 | COCO 2014 | 90% | [link]() |
57
+ | atc_base-coco-0.7 | COCO 2014 | 70% | [link]() |
58
+ | atc_base-coco-0.5 | COCO 2014 | 50% | [link]() |
59
+ | atc_base-coco-0.25 | COCO 2014 | 25% | [link]() |
60
+ | atc_small-coco-0.9 | COCO 2014 | 90% | [link]() |
61
+ | atc_small-coco-0.7 | COCO 2014 | 70% | [link]() |
62
+ | atc_small-coco-0.5 | COCO 2014 | 50% | [link]() |
63
+ | atc_small-coco-0.25 | COCO 2014 | 25% | [link]() |
64
+ | atc_tiny-coco-0.9 | COCO 2014 | 90% | [link]() |
65
+ | atc_tiny-coco-0.7 | COCO 2014 | 70% | [link]() |
66
+ | atc_tiny-coco-0.5 | COCO 2014 | 50% | [link]() |
67
+ | atc_tiny-coco-0.25 | COCO 2014 | 25% | [link]() |
68
+ <br>
69
+ | atc_base-nus-0.9 | NUS-WIDE | 90% | [link]() |
70
+ | atc_base-nus-0.7 | NUS-WIDE | 70% | [link]() |
71
+ | atc_base-nus-0.5 | NUS-WIDE | 50% | [link]() |
72
+ | atc_base-nus-0.25 | NUS-WIDE | 25% | [link]() |
73
+ | atc_small-nus-0.9 | NUS-WIDE | 90% | [link]() |
74
+ | atc_small-nus-0.7 | NUS-WIDE | 70% | [link]() |
75
+ | atc_small-nus-0.5 | NUS-WIDE | 50% | [link]() |
76
+ | atc_small-nus-0.25 | NUS-WIDE | 25% | [link]() |
77
+ | atc_tiny-nus-0.9 | NUS-WIDE | 90% | [link]() |
78
+ | atc_tiny-nus-0.7 | NUS-WIDE | 70% | [link]() |
79
+ | atc_tiny-nus-0.5 | NUS-WIDE | 50% | [link]() |
80
+ | atc_tiny-nus-0.25 | NUS-WIDE | 25% | [link]() |
81
+ <br>
82
+
83
+ Object Detection and Segmentation Models::
84
+ | Model Name | Dataset | Keep Rate | Weights |
85
+ |:-----------|:--------:|:-----------:|:--------:|
86
+ | atc_small-coco17-0.9 | COCO 2017 | 90% | [link]() |
87
+ | atc_small-coco17-0.7 | COCO 2017 | 70% | [link]() |
88
+ | atc_small-coco17-0.5 | COCO 2017 | 50% | [link]() |
89
+ | atc_small-coco17-0.25 | COCO 2017 | 25% | [link]() |
90
+ | tome_small-coco17-0.9 | COCO 2017 | 90% | [link]() |
91
+ | tome_small-coco17-0.7 | COCO 2017 | 70% | [link]() |
92
+ | tome_small-coco17-0.5 | COCO 2017 | 50% | [link]() |
93
+ | tome_small-coco17-0.25 | COCO 2017 | 25% | [link]() |
94
+ <br>
95
+ | atc_tiny-coco17-0.9 | COCO 2017 | 90% | [link]() |
96
+ | atc_tiny-coco17-0.7 | COCO 2017 | 70% | [link]() |
97
+ | atc_tiny-coco17-0.5 | COCO 2017 | 50% | [link]() |
98
+ | atc_tiny-coco17-0.25 | COCO 2017 | 25% | [link]() |
99
+ | tome_tiny-coco17-0.9 | COCO 2017 | 90% | [link]() |
100
+ | tome_tiny-coco17-0.7 | COCO 2017 | 70% | [link]() |
101
+ | tome_tiny-coco17-0.5 | COCO 2017 | 50% | [link]() |
102
+ | tome_tiny-coco17-0.25 | COCO 2017 | 25% | [link]() |
103
+ <br>