segformer-finetuned-Maize-10k-steps-sem

This model is a fine-tuned version of nvidia/mit-b5 on the koushikn/Maize_sem_seg dataset. It achieves the following results on the evaluation set:

• Loss: 0.0756
• Mean Iou: 0.9172
• Mean Accuracy: 0.9711
• Overall Accuracy: 0.9804
• Accuracy Background: 0.9834
• Accuracy Maize: 0.9588
• Iou Background: 0.9779
• Iou Maize: 0.8566

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: 0.001
• train_batch_size: 16
• eval_batch_size: 16
• seed: 1337
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: polynomial
• training_steps: 10000

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Background	Accuracy Maize	Iou Background	Iou Maize
0.0529	1.0	678	69.3785	0.4391	0.5	0.8781	1.0	0.0	0.8781	0.0
0.3755	2.0	1356	0.9455	0.4391	0.5	0.8781	1.0	0.0	0.8781	0.0
0.0603	3.0	2034	0.0920	0.8356	0.8602	0.9641	0.9976	0.7227	0.9607	0.7106
0.0341	4.0	2712	24.6203	0.4391	0.5	0.8781	1.0	0.0	0.8781	0.0
0.0332	5.0	3390	101.5635	0.4391	0.5	0.8781	1.0	0.0	0.8781	0.0
0.0331	6.0	4068	9.6824	0.4391	0.5	0.8781	1.0	0.0	0.8781	0.0
0.0302	7.0	4746	260.7923	0.4391	0.5	0.8781	1.0	0.0	0.8781	0.0
0.0305	8.0	5424	172.8153	0.4391	0.5	0.8781	1.0	0.0	0.8781	0.0
0.0313	9.0	6102	304.2714	0.4391	0.5	0.8781	1.0	0.0	0.8781	0.0
0.0301	10.0	6780	547.2355	0.4391	0.5	0.8781	1.0	0.0	0.8781	0.0
0.03	11.0	7458	224.2607	0.4391	0.5	0.8781	1.0	0.0	0.8781	0.0
0.0285	12.0	8136	116.3474	0.4391	0.5	0.8781	1.0	0.0	0.8781	0.0
0.0284	13.0	8814	96.8429	0.4391	0.5	0.8781	1.0	0.0	0.8781	0.0
0.0281	14.0	9492	54.2593	0.4391	0.5	0.8781	1.0	0.0	0.8781	0.0
0.028	14.75	10000	0.0756	0.9172	0.9711	0.9804	0.9834	0.9588	0.9779	0.8566

Framework versions

• Transformers 4.21.0.dev0
• Pytorch 1.10.0+cu102
• Datasets 2.3.2
• Tokenizers 0.12.1