segformer-b3-finetuned-IDD-L2

This model is a fine-tuned version of nvidia/mit-b3 on the IDD 10K Semantic Segmentation Dataset dataset. It achieves the following results on the evaluation set:

• Loss: 0.4239
• Mean Iou: 0.6943
• Mean Accuracy: 0.8029
• Overall Accuracy: 0.9050
• Accuracy Road: 0.9738
• Accuracy Parking: 0.7682
• Accuracy Sidewalk: 0.8188
• Accuracy Rail track: 0.6593
• Accuracy Person: 0.7127
• Accuracy Rider: 0.8413
• Accuracy Motorcycle: 0.8526
• Accuracy Autorickshaw: 0.9496
• Accuracy Truck: 0.9242
• Accuracy Curb: 0.7824
• Accuracy Fence: 0.5204
• Accuracy Billboard: 0.7477
• Accuracy Pole: 0.5379
• Accuracy Building: 0.8284
• Accuracy Vegetation: 0.9499
• Accuracy Sky: 0.9791
• Iou Road: 0.9405
• Iou Parking: 0.6269
• Iou Sidewalk: 0.6605
• Iou Rail track: 0.5518
• Iou Person: 0.5704
• Iou Rider: 0.7168
• Iou Motorcycle: 0.7206
• Iou Autorickshaw: 0.8859
• Iou Truck: 0.8723
• Iou Curb: 0.6553
• Iou Fence: 0.3869
• Iou Billboard: 0.5992
• Iou Pole: 0.4300
• Iou Building: 0.6570
• Iou Vegetation: 0.8699
• Iou Sky: 0.9644

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: 0.0006
• train_batch_size: 32
• eval_batch_size: 32
• 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	Accuracy Road	Accuracy Parking	Accuracy Sidewalk	Accuracy Rail track	Accuracy Person	Accuracy Rider	Accuracy Motorcycle	Accuracy Autorickshaw	Accuracy Truck	Accuracy Curb	Accuracy Fence	Accuracy Billboard	Accuracy Pole	Accuracy Building	Accuracy Vegetation	Accuracy Sky	Iou Road	Iou Parking	Iou Sidewalk	Iou Rail track	Iou Person	Iou Rider	Iou Motorcycle	Iou Autorickshaw	Iou Truck	Iou Curb	Iou Fence	Iou Billboard	Iou Pole	Iou Building	Iou Vegetation	Iou Sky
0.3533	1.0	216	0.3564	0.6076	0.7367	0.8778	0.9602	0.7496	0.7994	0.5899	0.4405	0.8377	0.6808	0.9400	0.8728	0.6738	0.4940	0.6091	0.4121	0.8063	0.9490	0.9713	0.9268	0.5838	0.5063	0.4737	0.3826	0.5677	0.5780	0.8134	0.7886	0.5744	0.2721	0.5245	0.3227	0.6052	0.8447	0.9567
0.3188	2.0	432	0.3370	0.6369	0.7517	0.8857	0.9524	0.7931	0.7515	0.5139	0.5994	0.7295	0.8174	0.9170	0.9412	0.7441	0.3820	0.6820	0.4824	0.7931	0.9488	0.9795	0.9246	0.5872	0.5761	0.4647	0.4696	0.6155	0.6574	0.8463	0.7979	0.5909	0.2944	0.5595	0.3701	0.6229	0.8530	0.9598
0.2794	3.0	648	0.3253	0.6489	0.7576	0.8906	0.9684	0.7253	0.6453	0.6555	0.6408	0.7941	0.7927	0.9455	0.8692	0.8066	0.4426	0.6544	0.4068	0.8332	0.9579	0.9823	0.9336	0.5970	0.5558	0.5219	0.5086	0.6485	0.6639	0.8440	0.8283	0.6121	0.3204	0.5678	0.3454	0.6273	0.8482	0.9594
0.2462	4.0	864	0.3544	0.6423	0.7616	0.8864	0.9800	0.6697	0.7797	0.6374	0.6485	0.8107	0.8088	0.9563	0.7332	0.7271	0.4623	0.7496	0.4739	0.8162	0.9588	0.9738	0.9319	0.5765	0.5879	0.5303	0.5129	0.6568	0.6676	0.7555	0.7060	0.6117	0.3227	0.5849	0.3841	0.6310	0.8573	0.9605
0.2229	5.0	1080	0.3165	0.6725	0.7888	0.8974	0.9755	0.6768	0.7878	0.7369	0.7024	0.7951	0.8411	0.9405	0.9081	0.7560	0.5121	0.7361	0.5060	0.8193	0.9428	0.9840	0.9344	0.5804	0.6349	0.5339	0.5384	0.6666	0.6856	0.8614	0.8525	0.6378	0.3646	0.5909	0.4025	0.6529	0.8626	0.9606
0.2032	6.0	1296	0.3236	0.6718	0.7807	0.8987	0.9685	0.7700	0.8081	0.6366	0.5997	0.8254	0.8517	0.9419	0.9085	0.7650	0.4252	0.7291	0.5105	0.8175	0.9581	0.9759	0.9359	0.6146	0.6512	0.5337	0.4996	0.6615	0.6898	0.8592	0.8460	0.6462	0.3397	0.5878	0.4050	0.6564	0.8601	0.9626
0.1934	7.0	1512	0.3410	0.6726	0.7842	0.8969	0.9745	0.7528	0.7805	0.6859	0.6589	0.8285	0.8226	0.9507	0.8671	0.7042	0.4820	0.7639	0.5195	0.8333	0.9432	0.9793	0.9373	0.6117	0.6513	0.5626	0.5307	0.6781	0.6820	0.8341	0.8262	0.6280	0.3689	0.5919	0.4098	0.6210	0.8660	0.9627
0.1785	8.0	1728	0.3354	0.6835	0.7995	0.9010	0.9718	0.7610	0.8231	0.6723	0.6939	0.8082	0.8543	0.9360	0.9260	0.7866	0.5626	0.7212	0.5049	0.8535	0.9344	0.9823	0.9374	0.6152	0.6373	0.5471	0.5562	0.6865	0.7034	0.8730	0.8577	0.6457	0.3876	0.5964	0.4102	0.6511	0.8686	0.9626
0.1574	9.0	1944	0.3406	0.6882	0.7984	0.9021	0.9695	0.7628	0.7986	0.6512	0.7484	0.8377	0.8298	0.9419	0.9062	0.7858	0.5157	0.7209	0.5417	0.8359	0.9523	0.9765	0.9349	0.6107	0.6552	0.5509	0.5639	0.7011	0.7099	0.8745	0.8585	0.6478	0.3888	0.5955	0.4312	0.6573	0.8682	0.9632
0.1532	10.0	2160	0.3501	0.6885	0.8016	0.9029	0.9710	0.7628	0.8349	0.6575	0.7238	0.8346	0.8534	0.9429	0.9152	0.8002	0.5157	0.7291	0.5361	0.8205	0.9477	0.9802	0.9379	0.6187	0.6408	0.5499	0.5777	0.7021	0.7123	0.8759	0.8626	0.6553	0.3800	0.5866	0.4265	0.6581	0.8686	0.9638
0.1473	11.0	2376	0.3508	0.6908	0.8031	0.9029	0.9679	0.7861	0.8132	0.6387	0.7161	0.8302	0.8583	0.9434	0.9251	0.7742	0.5477	0.7457	0.5671	0.8118	0.9463	0.9779	0.9373	0.6215	0.6421	0.5368	0.5738	0.7086	0.7170	0.8802	0.8635	0.6519	0.3959	0.5986	0.4351	0.6570	0.8699	0.9636
0.1326	12.0	2592	0.3717	0.6918	0.8014	0.9039	0.9796	0.7255	0.8101	0.6992	0.6947	0.8297	0.8605	0.9413	0.9169	0.7758	0.5590	0.7468	0.5254	0.8313	0.9489	0.9780	0.9393	0.6154	0.6607	0.5551	0.5720	0.7102	0.7118	0.8782	0.8669	0.6512	0.3937	0.5995	0.4227	0.6579	0.8701	0.9638
0.1279	13.0	2808	0.3743	0.6922	0.8017	0.9036	0.9707	0.7663	0.8159	0.6908	0.7243	0.8527	0.8349	0.9470	0.9234	0.7638	0.5047	0.7375	0.5337	0.8324	0.9513	0.9779	0.9387	0.6238	0.6646	0.5582	0.5797	0.7141	0.7159	0.8807	0.8691	0.6466	0.3789	0.5950	0.4251	0.6501	0.8707	0.9636
0.1283	14.0	3024	0.3879	0.6929	0.8008	0.9043	0.9761	0.7591	0.7942	0.6503	0.7241	0.8353	0.8515	0.9447	0.9212	0.7926	0.5276	0.7392	0.5363	0.8363	0.9466	0.9779	0.9398	0.6211	0.6540	0.5507	0.5813	0.7142	0.7159	0.8825	0.8683	0.6554	0.3850	0.6010	0.4277	0.6542	0.8709	0.9641
0.1202	15.0	3240	0.3912	0.6929	0.8027	0.9040	0.9723	0.7652	0.8185	0.6819	0.7089	0.8292	0.8588	0.9511	0.9173	0.7753	0.5258	0.7291	0.5432	0.8442	0.9446	0.9779	0.9402	0.6242	0.6535	0.5582	0.5784	0.7171	0.7160	0.8792	0.8697	0.6556	0.3825	0.5967	0.4284	0.6524	0.8706	0.9641
0.1132	16.0	3456	0.3924	0.6923	0.8011	0.9046	0.9724	0.7740	0.8158	0.6447	0.7160	0.8422	0.8486	0.9487	0.9222	0.7955	0.5001	0.7352	0.5417	0.8341	0.9463	0.9804	0.9396	0.6267	0.6582	0.5467	0.5689	0.7165	0.7166	0.8805	0.8663	0.6592	0.3722	0.6005	0.4308	0.6591	0.8706	0.9646
0.1132	17.0	3672	0.4103	0.6937	0.8026	0.9047	0.9736	0.7673	0.8237	0.6688	0.7178	0.8447	0.8440	0.9489	0.9228	0.7851	0.5073	0.7418	0.5356	0.8339	0.9463	0.9801	0.9399	0.6261	0.6605	0.5543	0.5730	0.7187	0.7193	0.8843	0.8701	0.6535	0.3812	0.6000	0.4281	0.6557	0.8704	0.9644
0.1105	18.0	3888	0.4118	0.6943	0.8037	0.9050	0.9729	0.7731	0.8208	0.6569	0.7078	0.8443	0.8494	0.9495	0.9250	0.7810	0.5338	0.7460	0.5457	0.8251	0.9469	0.9812	0.9403	0.6276	0.6607	0.5520	0.5714	0.7159	0.7194	0.8844	0.8705	0.6550	0.3891	0.5965	0.4332	0.6585	0.8706	0.9644
0.106	19.0	4104	0.4200	0.6947	0.8039	0.9050	0.9731	0.7706	0.8204	0.6548	0.7133	0.8435	0.8522	0.9503	0.9239	0.7830	0.5379	0.7400	0.5426	0.8286	0.9497	0.9785	0.9406	0.6276	0.6613	0.5508	0.5708	0.7165	0.7211	0.8857	0.8724	0.6555	0.3901	0.5990	0.4311	0.6579	0.8698	0.9644
0.1028	20.0	4320	0.4239	0.6943	0.8029	0.9050	0.9738	0.7682	0.8188	0.6593	0.7127	0.8413	0.8526	0.9496	0.9242	0.7824	0.5204	0.7477	0.5379	0.8284	0.9499	0.9791	0.9405	0.6269	0.6605	0.5518	0.5704	0.7168	0.7206	0.8859	0.8723	0.6553	0.3869	0.5992	0.4300	0.6570	0.8699	0.9644

Framework versions

• Transformers 4.42.4
• Pytorch 2.3.1+cu121
• Datasets 2.19.1
• Tokenizers 0.19.1