segformer-b2-finetuned-ade-512-512_corm Open Source Image Segmentation Model - Precise Handling of 512x512 Image Segmentation Tasks

Segformer B2 Finetuned Ade 512 512 Corm

Developed by mujerry

An image segmentation model based on the SegFormer architecture, pre-trained on the ADE20K dataset and further fine-tuned, excelling in 512x512 resolution image segmentation tasks

Image Segmentation

Transformers

Open Source License:Other #High-precision image segmentation #Plant pathology recognition #Agricultural visual inspection

Downloads 20

Release Time : 3/7/2025

Model Overview

This model is a variant of the SegFormer-B2 architecture, specifically designed for semantic image segmentation tasks. It demonstrates outstanding performance after fine-tuning on an unknown dataset, particularly achieving high accuracy in segmenting bulbs and damaged areas

Model Features

High-precision segmentation

Achieves a mean Intersection over Union (mIoU) of 0.9264 and mean accuracy of 0.9599 on the evaluation set

Multi-class recognition

Accurately distinguishes between three target classes: background, bulbs, and damaged areas

Optimized training

Utilizes linear learning rate scheduling and warm-up strategy for 40 epochs of fine-tuning

Model Capabilities

Semantic image segmentation

Multi-class pixel-level classification

High-resolution image processing (512x512)

Use Cases

Agricultural image analysis

Crop disease detection

Identifies and segments diseased areas in crop images

Damage area segmentation accuracy reaches 94.56%

Botanical research

Plant organ analysis

Precisely segments plant organ structures such as bulbs

Bulb segmentation accuracy reaches 93.62%

🚀 segformer-b2-finetuned-ade-512-512_corm

This model is a fine - tuned version of nvidia/segformer-b2-finetuned-ade-512-512, designed for image segmentation tasks with high accuracy.

🚀 Quick Start

This model is a fine - tuned version of nvidia/segformer-b2-finetuned-ade-512-512 on the None dataset. It achieves the following results on the evaluation set:

Loss: 0.0415
Mean Iou: 0.9264
Mean Accuracy: 0.9599
Overall Accuracy: 0.9860
Accuracy Background: 0.9978
Accuracy Corm: 0.9362
Accuracy Damage: 0.9456
Iou Background: 0.9942
Iou Corm: 0.8799
Iou Damage: 0.9052

📚 Documentation

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 6e - 05
train_batch_size: 8
eval_batch_size: 8
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon = 1e - 08
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.05
num_epochs: 40

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Background	Accuracy Corm	Accuracy Damage	Iou Background	Iou Corm	Iou Damage
0.8746	0.9524	20	0.8170	0.4637	0.6489	0.8404	0.9133	0.0390	0.9944	0.9132	0.0327	0.4451
0.61	1.9048	40	0.4500	0.7608	0.8748	0.9451	0.9731	0.6946	0.9566	0.9730	0.6023	0.7071
0.3681	2.8571	60	0.2802	0.8597	0.9314	0.9711	0.9879	0.8621	0.9443	0.9873	0.7716	0.8201
0.2433	3.8095	80	0.2201	0.8866	0.9456	0.9774	0.9916	0.9159	0.9293	0.9904	0.8181	0.8513
0.1669	4.7619	100	0.1607	0.8891	0.9431	0.9783	0.9930	0.8723	0.9640	0.9915	0.8178	0.8580
0.1484	5.7143	120	0.1250	0.9017	0.9489	0.9812	0.9964	0.9423	0.9080	0.9932	0.8432	0.8688
0.1126	6.6667	140	0.1024	0.9092	0.9524	0.9827	0.9962	0.9247	0.9363	0.9934	0.8539	0.8804
0.0909	7.6190	160	0.0932	0.9017	0.9492	0.9813	0.9968	0.9563	0.8944	0.9935	0.8445	0.8670
0.0994	8.5714	180	0.0803	0.9122	0.9527	0.9833	0.9967	0.9118	0.9495	0.9936	0.8568	0.8861
0.0768	9.5238	200	0.0716	0.9147	0.9533	0.9838	0.9975	0.9247	0.9376	0.9937	0.8615	0.8889
0.0749	10.4762	220	0.0671	0.9177	0.9550	0.9844	0.9973	0.9191	0.9487	0.9939	0.8661	0.8932
0.0663	11.4286	240	0.0668	0.9097	0.9528	0.9829	0.9973	0.9528	0.9083	0.9939	0.8558	0.8795
0.0725	12.3810	260	0.0608	0.9189	0.9554	0.9847	0.9974	0.9123	0.9564	0.9940	0.8677	0.8951
0.0594	13.3333	280	0.0588	0.9167	0.9533	0.9843	0.9975	0.9000	0.9625	0.9940	0.8622	0.8940
0.062	14.2857	300	0.0552	0.9201	0.9565	0.9849	0.9972	0.9170	0.9553	0.9941	0.8691	0.8970
0.0535	15.2381	320	0.0543	0.9195	0.9559	0.9848	0.9972	0.9078	0.9626	0.9942	0.8683	0.8962
0.0555	16.1905	340	0.0517	0.9212	0.9566	0.9851	0.9973	0.9113	0.9612	0.9942	0.8704	0.8990
0.0553	17.1429	360	0.0513	0.9198	0.9553	0.9849	0.9975	0.9047	0.9638	0.9942	0.8679	0.8974
0.0572	18.0952	380	0.0501	0.9219	0.9563	0.9853	0.9977	0.9108	0.9603	0.9942	0.8713	0.9002
0.0503	19.0476	400	0.0483	0.9245	0.9573	0.9856	0.9981	0.9212	0.9525	0.9940	0.8757	0.9037
0.0539	20.0	420	0.0474	0.9245	0.9593	0.9857	0.9974	0.9309	0.9497	0.9942	0.8769	0.9024
0.0542	20.9524	440	0.0484	0.9202	0.9575	0.9849	0.9978	0.9511	0.9235	0.9941	0.8718	0.8949
0.033	21.9048	460	0.0478	0.9209	0.9576	0.9850	0.9977	0.9464	0.9287	0.9941	0.8726	0.8961
0.0421	22.8571	480	0.0452	0.9247	0.9591	0.9857	0.9974	0.9244	0.9555	0.9942	0.8766	0.9033
0.0472	23.8095	500	0.0455	0.9243	0.9583	0.9857	0.9976	0.9231	0.9543	0.9942	0.8759	0.9028
0.0381	24.7619	520	0.0456	0.9233	0.9570	0.9855	0.9977	0.9109	0.9625	0.9942	0.8732	0.9026
0.0486	25.7143	540	0.0444	0.9249	0.9593	0.9857	0.9978	0.9408	0.9394	0.9941	0.8780	0.9026
0.0501	26.6667	560	0.0458	0.9208	0.9579	0.9850	0.9977	0.9508	0.9252	0.9942	0.8725	0.8957
0.0343	27.6190	580	0.0436	0.9251	0.9594	0.9857	0.9978	0.9413	0.9391	0.9941	0.8782	0.9031
0.0407	28.5714	600	0.0434	0.9251	0.9597	0.9858	0.9977	0.9416	0.9396	0.9942	0.8784	0.9028
0.0419	29.5238	620	0.0445	0.9221	0.9586	0.9852	0.9977	0.9496	0.9285	0.9942	0.8743	0.8978
0.0506	30.4762	640	0.0425	0.9262	0.9593	0.9860	0.9978	0.9311	0.9491	0.9942	0.8791	0.9053
0.0422	31.4286	660	0.0424	0.9262	0.9595	0.9860	0.9977	0.9267	0.9540	0.9942	0.8790	0.9054
0.0362	32.3810	680	0.0425	0.9258	0.9600	0.9859	0.9977	0.9402	0.9421	0.9942	0.8793	0.9039
0.0437	33.3333	700	0.0424	0.9262	0.9599	0.9860	0.9978	0.9377	0.9441	0.9942	0.8796	0.9047
0.0363	34.2857	720	0.0415	0.9264	0.9602	0.9860	0.9976	0.9367	0.9463	0.9942	0.8800	0.9049
0.039	35.2381	740	0.0421	0.9267	0.9596	0.9861	0.9978	0.9290	0.9521	0.9942	0.8798	0.9060
0.0425	36.1905	760	0.0418	0.9259	0.9598	0.9859	0.9978	0.9391	0.9426	0.9942	0.8794	0.9040
0.0462	37.1429	780	0.0417	0.9267	0.9600	0.9861	0.9976	0.9311	0.9513	0.9942	0.8801	0.9057
0.0466	38.0952	800	0.0416	0.9261	0.9599	0.9860	0.9978	0.9392	0.9427	0.9942	0.8795	0.9045
0.0428	39.0476	820	0.0414	0.9266	0.9598	0.9861	0.9978	0.9323	0.9494	0.9942	0.8800	0.9057
0.04	40.0	840	0.0415	0.9264	0.9599	0.9860	0.9978	0.9362	0.9456	0.9942	0.8799	0.9052

Framework versions

Transformers 4.44.1
Pytorch 2.6.0+cpu
Datasets 2.21.0
Tokenizers 0.19.1

📄 License

License: other

🔧 Technical Details

Property	Details
Model Type	segformer-b2-finetuned-ade-512-512_corm
Base Model	nvidia/segformer-b2-finetuned-ade-512-512
Tags	vision, image-segmentation, generated_from_trainer

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご