segformer-b0-scene-parse-150-20
This model is a fine-tuned version of klentree/segformer-b0-scene-parse-150 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 0.9683
- Mean Iou: 0.2488
- Mean Accuracy: 0.6439
- Overall Accuracy: 0.4669
- Per Category Iou: [0.044136660707036465, 0.4534135035032552]
- Per Category Accuracy: [0.8321388577827548, 0.45573371779771177]
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: 6e-05
- train_batch_size: 1
- eval_batch_size: 1
- seed: 42
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 20
Training results
Training Loss | Epoch | Step | Validation Loss | Mean Iou | Mean Accuracy | Overall Accuracy | Per Category Iou | Per Category Accuracy |
---|---|---|---|---|---|---|---|---|
No log | 1.0 | 133 | 0.1391 | 0.5047 | 0.5218 | 0.9675 | [0.04196317239808363, 0.967443434576684] | [0.04813532865044623, 0.9955161397039883] |
No log | 2.0 | 266 | 0.1449 | 0.5154 | 0.5451 | 0.9597 | [0.07117236629542235, 0.9595300984781853] | [0.10450134039159795, 0.9857244117018465] |
No log | 3.0 | 399 | 0.1669 | 0.5128 | 0.5453 | 0.9569 | [0.06888629524029359, 0.9567357547653306] | [0.10785231938647392, 0.9827560505657416] |
0.3733 | 4.0 | 532 | 0.2122 | 0.5112 | 0.5576 | 0.9478 | [0.07483006503985525, 0.9475505125490813] | [0.14279581933557298, 0.9723116234544236] |
0.3733 | 5.0 | 665 | 0.2619 | 0.5036 | 0.6005 | 0.9199 | [0.0879079186151223, 0.9192549826254068] | [0.2610336285588245, 0.9399632680244574] |
0.3733 | 6.0 | 798 | 0.3013 | 0.4815 | 0.6494 | 0.8758 | [0.08872990969490727, 0.8742854201811897] | [0.40877362652278665, 0.8900430360382596] |
0.3733 | 7.0 | 931 | 0.3093 | 0.4823 | 0.6044 | 0.8905 | [0.0750589475389164, 0.8895250315242322] | [0.30035460992907803, 0.9084973158596108] |
0.2604 | 8.0 | 1064 | 0.2087 | 0.5090 | 0.5805 | 0.9342 | [0.0842029042177912, 0.9337687360922476] | [0.20460314228511317, 0.9564103131735049] |
0.2604 | 9.0 | 1197 | 0.4990 | 0.4044 | 0.6864 | 0.7459 | [0.06763197122282019, 0.7411177869667952] | [0.6230818826563508, 0.7496334354079776] |
0.2604 | 10.0 | 1330 | 0.3745 | 0.4472 | 0.6666 | 0.8206 | [0.07657083870114431, 0.8178660489858295] | [0.5029200176456615, 0.830259490697111] |
0.2604 | 11.0 | 1463 | 1.2307 | 0.1467 | 0.5701 | 0.2778 | [0.03481955130980837, 0.25848886923646197] | [0.8806949675930639, 0.2594289909506597] |
0.1761 | 12.0 | 1596 | 0.2577 | 0.4905 | 0.6244 | 0.8952 | [0.0867593572032901, 0.8941645624438329] | [0.33649258542875565, 0.9122507365946599] |
0.1761 | 13.0 | 1729 | 0.4025 | 0.4353 | 0.6588 | 0.8030 | [0.07054198842750235, 0.7999897274079024] | [0.505458278190641, 0.8120503803644273] |
0.1761 | 14.0 | 1862 | 5.8258 | 0.0188 | 0.5025 | 0.0372 | [0.029723375945066684, 0.007894678045920561] | [0.9970477450880586, 0.007895388558467544] |
0.1761 | 15.0 | 1995 | 0.7888 | 0.3005 | 0.6622 | 0.5618 | [0.049354828101168945, 0.5516162745499266] | [0.7689979300281652, 0.5555007860940694] |
0.1331 | 16.0 | 2128 | 0.3346 | 0.4627 | 0.6483 | 0.8488 | [0.07848367609919223, 0.846876325384084] | [0.43516067732193153, 0.8614587046651099] |
0.1331 | 17.0 | 2261 | 2.7924 | 0.0254 | 0.5076 | 0.0497 | [0.03002291425841882, 0.02085235460380035] | [0.9943313312294275, 0.0208559580633969] |
0.1331 | 18.0 | 2394 | 1.5711 | 0.1377 | 0.5818 | 0.2605 | [0.03562331490097667, 0.23971062459649647] | [0.9234127388102752, 0.24027028849931653] |
0.1234 | 19.0 | 2527 | 0.2332 | 0.4998 | 0.6022 | 0.9146 | [0.08556669218462386, 0.9139353359496684] | [0.2700549730224982, 0.9342724393702061] |
0.1234 | 20.0 | 2660 | 0.9683 | 0.2488 | 0.6439 | 0.4669 | [0.044136660707036465, 0.4534135035032552] | [0.8321388577827548, 0.45573371779771177] |
Framework versions
- Transformers 4.37.0
- Pytorch 2.1.0+cu121
- Datasets 2.16.1
- Tokenizers 0.15.0
- Downloads last month
- 0
Unable to determine this model’s pipeline type. Check the
docs
.