Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,62 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
datasets:
|
| 3 |
+
- hieupth/cad
|
| 4 |
+
- shreyanshu09/Block_Diagram
|
| 5 |
+
- krowiemlekommm/PJN_CHARTS
|
| 6 |
+
- corto-ai/handwritten-text
|
| 7 |
+
- katanaml-org/invoices-donut-data-v1
|
| 8 |
+
- mathieu1256/FATURA2-invoices
|
| 9 |
+
- AjitRawat/invoice
|
| 10 |
+
- HuggingFaceM4/DocumentVQA
|
| 11 |
+
- ajaynmopidevi/DocumentIDEFICS_QA
|
| 12 |
+
- daitavan/financial-documents
|
| 13 |
+
- Anas989898/Vision-OCR-Financial-Reports-10k
|
| 14 |
+
- dpdl-benchmark/places100-easy
|
| 15 |
+
- xirigh/people
|
| 16 |
+
- huggan/flowers-102-categories
|
| 17 |
+
- iamkaikai/amazing_logos
|
| 18 |
+
- iamkaikai/amazing_logos_v2
|
| 19 |
+
- manelreghima/companies_logos
|
| 20 |
+
- samp3209/logo-dataset
|
| 21 |
+
- salmonhumorous/logo-blip-caption
|
| 22 |
+
- dream-textures/textures-color-1k
|
| 23 |
+
- GATE-engine/describable_textures
|
| 24 |
+
- ppierzc/ios-app-icons
|
| 25 |
+
- Lancelot53/android_icon_dataset
|
| 26 |
+
- naxalpha/stable-icons-128
|
| 27 |
+
- yaneivan/memes_caption
|
| 28 |
+
- Bruece/office-home-clipart-caption
|
| 29 |
+
base_model:
|
| 30 |
+
- timm/resnext50_32x4d.fb_swsl_ig1b_ft_in1k
|
| 31 |
+
---
|
| 32 |
+
# fine_tuned_image_relevance_model
|
| 33 |
+
This model is a fine-tuned version of [resnext50_32x4d.fb_swsl_ig1b_ft_in1k](https://huggingface.co/timm/resnext50_32x4d.fb_swsl_ig1b_ft_in1k) on an aggregated dataset of images that were classified as relevant (1.0) or irrelevant (0.0). It achieves the following results on the validation set:
|
| 34 |
+
|
| 35 |
+
- Loss: 0.1032
|
| 36 |
+
- Accuracy: 0.9936
|
| 37 |
+
|
| 38 |
+
## Training hyperparameters
|
| 39 |
+
The following hyperparameters were used during training:
|
| 40 |
+
|
| 41 |
+
- learning_rate: 4e-06
|
| 42 |
+
- train_batch_size: 8
|
| 43 |
+
- valid_batch_size: 8
|
| 44 |
+
- seed: seed not explicitly set
|
| 45 |
+
- optimizer: torch.optim.AdamW(resnet_model.parameters(), lr=lr, eps=0.000001)
|
| 46 |
+
- lr_scheduler_type: OneCycleLR
|
| 47 |
+
- num_epochs: 6
|
| 48 |
+
|
| 49 |
+
## Training results
|
| 50 |
+
| Training Loss | Epoch | Validation Loss | Accuracy |
|
| 51 |
+
| :-----------: | :---: | :-------------: | :------: |
|
| 52 |
+
| 0.5536 | 1 | 0.3270 | 0.9856 |
|
| 53 |
+
| 0.3176 | 2 | 0.1720 | 0.9922 |
|
| 54 |
+
| 0.1887 | 3 | 0.1332 | 0.9944 |
|
| 55 |
+
| 0.1280 | 4 | 0.1146 | 0.9938 |
|
| 56 |
+
| 0.1116 | 5 | 0.1236 | 0.9938 |
|
| 57 |
+
| 0.1016 | 6 | 0.1032 | 0.9936 |
|
| 58 |
+
|
| 59 |
+
## Framework versions
|
| 60 |
+
- timm 1.0.19
|
| 61 |
+
- PyTorch 2.8.0+cpu
|
| 62 |
+
- Datasets 4.0.0
|