efederici commited on
Commit
b16e21a
1 Parent(s): 74d3a36

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +28 -30
README.md CHANGED
@@ -3,7 +3,6 @@ license: apache-2.0
3
  tags:
4
  - image-classification
5
  - vision
6
- - generated_from_trainer
7
  datasets:
8
  - cats_vs_dogs
9
  metrics:
@@ -24,9 +23,6 @@ model-index:
24
  value: 0.9973333333333333
25
  ---
26
 
27
- <!-- This model card has been generated automatically according to the information the Trainer had access to. You
28
- should probably proofread and complete it, then remove this comment. -->
29
-
30
  # convnext-base-224-22k-1k-orig-cats-vs-dogs
31
 
32
  This model is a fine-tuned version of [facebook/convnext-base-224-22k-1k](https://huggingface.co/facebook/convnext-base-224-22k-1k) on the cats_vs_dogs dataset.
@@ -34,19 +30,15 @@ It achieves the following results on the evaluation set:
34
  - Loss: 0.0103
35
  - Accuracy: 0.9973
36
 
37
- ## Model description
38
-
39
- More information needed
40
-
41
- ## Intended uses & limitations
42
-
43
- More information needed
44
 
45
- ## Training and evaluation data
46
 
47
- More information needed
48
 
49
- ## Training procedure
50
 
51
  ### Training hyperparameters
52
 
@@ -59,20 +51,26 @@ The following hyperparameters were used during training:
59
  - lr_scheduler_type: linear
60
  - num_epochs: 5.0
61
 
62
- ### Training results
63
-
64
- | Training Loss | Epoch | Step | Validation Loss | Accuracy |
65
- |:-------------:|:-----:|:----:|:---------------:|:--------:|
66
- | 0.0072 | 1.0 | 664 | 0.0140 | 0.9973 |
67
- | 0.0062 | 2.0 | 1328 | 0.0111 | 0.9965 |
68
- | 0.0311 | 3.0 | 1992 | 0.0115 | 0.9976 |
69
- | 0.0021 | 4.0 | 2656 | 0.0122 | 0.9973 |
70
- | 0.0248 | 5.0 | 3320 | 0.0103 | 0.9973 |
71
-
72
-
73
- ### Framework versions
74
 
75
- - Transformers 4.20.0
76
- - Pytorch 1.11.0+cu113
77
- - Datasets 2.3.3.dev0
78
- - Tokenizers 0.12.1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
3
  tags:
4
  - image-classification
5
  - vision
 
6
  datasets:
7
  - cats_vs_dogs
8
  metrics:
 
23
  value: 0.9973333333333333
24
  ---
25
 
 
 
 
26
  # convnext-base-224-22k-1k-orig-cats-vs-dogs
27
 
28
  This model is a fine-tuned version of [facebook/convnext-base-224-22k-1k](https://huggingface.co/facebook/convnext-base-224-22k-1k) on the cats_vs_dogs dataset.
 
30
  - Loss: 0.0103
31
  - Accuracy: 0.9973
32
 
33
+ <p align="center">
34
+ <img src="https://files.ocula.com/anzax/09/09f77133-7740-4130-a567-84fb56736362_650_544.jpg" width="600"> </br>
35
+ Jockum Nordström, Cat Dog Cat, 2016
36
+ </p>
 
 
 
37
 
38
+ ## Model description
39
 
40
+ The "Roaring 20s" of visual recognition began with the introduction of Vision Transformers (ViTs), which quickly superseded ConvNets as the state-of-the-art image classification model. A vanilla ViT, on the other hand, faces difficulties when applied to general computer vision tasks such as object detection and semantic segmentation. It is the hierarchical Transformers (e.g., Swin Transformers) that reintroduced several ConvNet priors, making Transformers practically viable as a generic vision backbone and demonstrating remarkable performance on a wide variety of vision tasks. However, the effectiveness of such hybrid approaches is still largely credited to the intrinsic superiority of Transformers, rather than the inherent inductive biases of convolutions. In this work, we reexamine the design spaces and test the limits of what a pure ConvNet can achieve. We gradually "modernize" a standard ResNet toward the design of a vision Transformer, and discover several key components that contribute to the performance difference along the way. The outcome of this exploration is a family of pure ConvNet models dubbed ConvNeXt. Constructed entirely from standard ConvNet modules, ConvNeXts compete favorably with Transformers in terms of accuracy and scalability, achieving 87.8% ImageNet top-1 accuracy and outperforming Swin Transformers on COCO detection and ADE20K segmentation, while maintaining the simplicity and efficiency of standard ConvNets.
41
 
 
42
 
43
  ### Training hyperparameters
44
 
 
51
  - lr_scheduler_type: linear
52
  - num_epochs: 5.0
53
 
 
 
 
 
 
 
 
 
 
 
 
 
54
 
55
+ ### BibTeX entry and citation info
56
+
57
+ ```bibtex
58
+ @article{DBLP:journals/corr/abs-2201-03545,
59
+ author = {Zhuang Liu and
60
+ Hanzi Mao and
61
+ Chao{-}Yuan Wu and
62
+ Christoph Feichtenhofer and
63
+ Trevor Darrell and
64
+ Saining Xie},
65
+ title = {A ConvNet for the 2020s},
66
+ journal = {CoRR},
67
+ volume = {abs/2201.03545},
68
+ year = {2022},
69
+ url = {https://arxiv.org/abs/2201.03545},
70
+ eprinttype = {arXiv},
71
+ eprint = {2201.03545},
72
+ timestamp = {Thu, 20 Jan 2022 14:21:35 +0100},
73
+ biburl = {https://dblp.org/rec/journals/corr/abs-2201-03545.bib},
74
+ bibsource = {dblp computer science bibliography, https://dblp.org}
75
+ }
76
+ ```