Edit model card

swinv2-large-panorama-IQA

This model is a fine-tuned version of microsoft/swinv2-large-patch4-window12to16-192to256-22kto1k-ft on the isiqa-2019-hf dataset. It achieves the following results on the evaluation set:

  • Loss: 0.0352
  • Srocc: 0.0683
  • Lcc: 0.1820

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 1e-05
  • train_batch_size: 4
  • eval_batch_size: 32
  • seed: 10
  • gradient_accumulation_steps: 16
  • total_train_batch_size: 64
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_ratio: 0.1
  • num_epochs: 50.0

Training results

Training Loss Epoch Step Validation Loss Srocc Lcc
No log 0.8889 3 0.4401 -0.1586 -0.1147
No log 1.7778 6 0.2206 -0.2675 -0.1545
0.3084 2.9630 10 0.1910 -0.2754 -0.1813
0.3084 3.8519 13 0.2334 -0.2170 -0.1511
0.3084 4.7407 16 0.1484 -0.2 -0.1310
0.0852 5.9259 20 0.1259 -0.1021 -0.0852
0.0852 6.8148 23 0.1552 -0.0709 -0.0595
0.0852 8.0 27 0.0942 -0.0948 -0.0584
0.0406 8.8889 30 0.0841 -0.0480 -0.0550
0.0406 9.7778 33 0.0886 -0.0576 -0.0448
0.0406 10.9630 37 0.0721 -0.0773 -0.0474
0.023 11.8519 40 0.0697 -0.0446 -0.0364
0.023 12.7407 43 0.0577 -0.0217 -0.0091
0.023 13.9259 47 0.0666 -0.0314 0.0112
0.0136 14.8148 50 0.0525 -0.0501 0.0060
0.0136 16.0 54 0.0626 -0.0178 0.0504
0.0136 16.8889 57 0.0438 0.0159 0.0827
0.0113 17.7778 60 0.0503 0.0741 0.1074
0.0113 18.9630 64 0.0429 0.0818 0.1129
0.0113 19.8519 67 0.0455 0.0874 0.1188
0.0097 20.7407 70 0.0597 0.0926 0.1316
0.0097 21.9259 74 0.0397 0.0614 0.1446
0.0097 22.8148 77 0.0529 0.0778 0.1637
0.0084 24.0 81 0.0366 0.0716 0.1761
0.0084 24.8889 84 0.0352 0.0683 0.1820
0.0084 25.7778 87 0.0491 0.0970 0.1848
0.0078 26.9630 91 0.0396 0.0984 0.1831
0.0078 27.8519 94 0.0395 0.1012 0.1856
0.0078 28.7407 97 0.0426 0.1097 0.1956
0.0063 29.9259 101 0.0370 0.1002 0.1984

Framework versions

  • Transformers 4.42.3
  • Pytorch 2.1.2
  • Datasets 2.20.0
  • Tokenizers 0.19.1
Downloads last month
5
Safetensors
Model size
195M params
Tensor type
F32
·
Inference Examples
Unable to determine this model's library. Check the docs .

Model tree for DiTo97/swinv2-large-panorama-IQA