segformer-b4-finetuned-IDD-L2_v2

This model is a fine-tuned version of nvidia/mit-b4 on the IDD 20K Semantic Segmentation Dataset dataset. It achieves the following results on the evaluation set:

• Loss: 0.4199
• Mean Iou: 0.7150
• Mean Accuracy: 0.8193
• Overall Accuracy: 0.9075
• Accuracy Road: 0.9704
• Accuracy Parking: 0.7964
• Accuracy Sidewalk: 0.8037
• Accuracy Rail track: 0.6379
• Accuracy Person: 0.8041
• Accuracy Rider: 0.8391
• Accuracy Motorcycle: 0.8711
• Accuracy Autorickshaw: 0.9431
• Accuracy Truck: 0.9116
• Accuracy Curb: 0.8175
• Accuracy Fence: 0.5566
• Accuracy Billboard: 0.7593
• Accuracy Pole: 0.5995
• Accuracy Building: 0.8647
• Accuracy Vegetation: 0.9506
• Accuracy Sky: 0.9837
• Iou Road: 0.9380
• Iou Parking: 0.6587
• Iou Sidewalk: 0.6591
• Iou Rail track: 0.5041
• Iou Person: 0.6800
• Iou Rider: 0.7182
• Iou Motorcycle: 0.7558
• Iou Autorickshaw: 0.8824
• Iou Truck: 0.8412
• Iou Curb: 0.6810
• Iou Fence: 0.4420
• Iou Billboard: 0.6199
• Iou Pole: 0.4787
• Iou Building: 0.7301
• Iou Vegetation: 0.8798
• Iou Sky: 0.9704

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
0.3232	1.0	403	0.3421	0.6363	0.7588	0.8789	0.9456	0.7825	0.6804	0.5401	0.7444	0.7825	0.7971	0.8830	0.9010	0.7907	0.4003	0.6603	0.4316	0.8796	0.9514	0.9711	0.9181	0.6123	0.5711	0.4297	0.5641	0.6018	0.6655	0.7989	0.7427	0.5943	0.3208	0.5247	0.3546	0.6657	0.8550	0.9614
0.2743	2.0	806	0.3112	0.6670	0.7944	0.8897	0.9468	0.8278	0.8108	0.5271	0.7564	0.8043	0.8322	0.9343	0.8851	0.7830	0.5247	0.7534	0.5963	0.8113	0.9392	0.9774	0.9211	0.6279	0.5911	0.4497	0.5987	0.6241	0.6907	0.8423	0.7847	0.6343	0.3785	0.5673	0.4378	0.6921	0.8670	0.9654
0.2627	3.0	1209	0.2994	0.6779	0.7857	0.8951	0.9717	0.7678	0.7806	0.6031	0.7208	0.7611	0.7792	0.9423	0.8692	0.7819	0.5793	0.6944	0.5309	0.8538	0.9555	0.9794	0.9331	0.6395	0.6478	0.4797	0.6203	0.6333	0.6912	0.8494	0.8087	0.6316	0.3783	0.5788	0.4226	0.7052	0.8611	0.9660
0.2527	4.0	1612	0.3124	0.6736	0.7825	0.8932	0.9673	0.8051	0.8018	0.5394	0.7225	0.7694	0.8487	0.9610	0.7878	0.7202	0.5027	0.7432	0.5638	0.8606	0.9453	0.9808	0.9318	0.6447	0.6291	0.4625	0.6229	0.6499	0.7082	0.7923	0.7447	0.6308	0.4085	0.5863	0.4323	0.6961	0.8705	0.9671
0.2129	5.0	2015	0.3007	0.6909	0.8002	0.8993	0.9617	0.8104	0.8030	0.5452	0.8135	0.8201	0.8506	0.9398	0.8892	0.7783	0.4935	0.7488	0.5243	0.8880	0.9526	0.9837	0.9332	0.6515	0.6486	0.4671	0.6401	0.6730	0.7200	0.8672	0.8252	0.6531	0.4073	0.5976	0.4299	0.7026	0.8705	0.9676
0.2095	6.0	2418	0.3021	0.6959	0.7995	0.9008	0.9626	0.8123	0.7592	0.6501	0.7396	0.8123	0.8521	0.9405	0.9022	0.7873	0.5159	0.7070	0.5396	0.8733	0.9550	0.9835	0.9326	0.6431	0.6494	0.4957	0.6463	0.6757	0.7275	0.8611	0.8230	0.6597	0.4194	0.5993	0.4457	0.7180	0.8698	0.9679
0.1906	7.0	2821	0.3018	0.6986	0.8095	0.9018	0.9677	0.7849	0.7730	0.6484	0.7610	0.8164	0.8760	0.9270	0.9075	0.7919	0.5954	0.7476	0.5649	0.8528	0.9527	0.9844	0.9345	0.6475	0.6453	0.4947	0.6535	0.6774	0.7330	0.8712	0.8289	0.6530	0.4239	0.5989	0.4555	0.7184	0.8731	0.9683
0.1757	8.0	3224	0.3220	0.6988	0.8086	0.9022	0.9709	0.8020	0.8179	0.6339	0.7666	0.8013	0.8672	0.9300	0.9026	0.8163	0.5430	0.7349	0.5659	0.8597	0.9446	0.9815	0.9375	0.6562	0.6483	0.5014	0.6540	0.6825	0.7385	0.8605	0.8027	0.6652	0.4193	0.6021	0.4517	0.7154	0.8760	0.9690
0.1609	9.0	3627	0.3290	0.7031	0.8145	0.9020	0.9576	0.8167	0.8228	0.6573	0.7885	0.8251	0.8611	0.9455	0.8984	0.8148	0.5296	0.7482	0.5730	0.8655	0.9438	0.9845	0.9309	0.6443	0.6498	0.5032	0.6633	0.6978	0.7419	0.8697	0.8278	0.6664	0.4238	0.6099	0.4607	0.7147	0.8766	0.9694
0.1455	10.0	4030	0.3234	0.7075	0.8158	0.9051	0.9709	0.7877	0.8051	0.6510	0.7843	0.8262	0.8581	0.9397	0.9139	0.8015	0.5653	0.7671	0.5900	0.8619	0.9454	0.9842	0.9375	0.6576	0.6413	0.5001	0.6722	0.7057	0.7459	0.8741	0.8379	0.6710	0.4329	0.6055	0.4692	0.7216	0.8786	0.9696
0.1442	11.0	4433	0.3395	0.7110	0.8123	0.9057	0.9701	0.7943	0.7823	0.6195	0.7953	0.8308	0.8577	0.9463	0.9047	0.8048	0.5494	0.7363	0.6055	0.8693	0.9459	0.9841	0.9369	0.6520	0.6611	0.4967	0.6758	0.7119	0.7499	0.8777	0.8401	0.6752	0.4414	0.6124	0.4731	0.7225	0.8801	0.9695
0.1312	12.0	4836	0.3508	0.7094	0.8143	0.9052	0.9698	0.7827	0.8066	0.6213	0.7915	0.8336	0.8708	0.9370	0.9100	0.8015	0.5581	0.7499	0.5969	0.8652	0.9509	0.9836	0.9365	0.6492	0.6539	0.4960	0.6763	0.7095	0.7501	0.8767	0.8344	0.6771	0.4347	0.6143	0.4698	0.7251	0.8769	0.9698
0.1302	13.0	5239	0.3644	0.7119	0.8162	0.9063	0.9713	0.7738	0.8040	0.6611	0.7956	0.8497	0.8647	0.9418	0.9090	0.8067	0.5416	0.7426	0.5875	0.8732	0.9526	0.9831	0.9377	0.6528	0.6609	0.5044	0.6768	0.7091	0.7507	0.8791	0.8385	0.6829	0.4343	0.6197	0.4711	0.7236	0.8783	0.9699
0.1252	14.0	5642	0.3692	0.7105	0.8179	0.9057	0.9724	0.7785	0.8000	0.6623	0.8099	0.8213	0.8761	0.9434	0.8984	0.8075	0.5618	0.7751	0.5945	0.8508	0.9493	0.9851	0.9381	0.6544	0.6539	0.5028	0.6744	0.7074	0.7503	0.8758	0.8317	0.6760	0.4457	0.6159	0.4716	0.7201	0.8795	0.9698
0.1179	15.0	6045	0.3840	0.7141	0.8184	0.9070	0.9706	0.7989	0.8076	0.6553	0.7914	0.8402	0.8607	0.9431	0.9118	0.8123	0.5572	0.7534	0.5840	0.8765	0.9474	0.9831	0.9380	0.6597	0.6589	0.5057	0.6784	0.7124	0.7551	0.8787	0.8391	0.6866	0.4467	0.6210	0.4728	0.7225	0.8802	0.9704
0.1149	16.0	6448	0.3896	0.7134	0.8183	0.9071	0.9715	0.7917	0.8146	0.6380	0.8009	0.8285	0.8722	0.9430	0.9112	0.8158	0.5468	0.7617	0.5978	0.8662	0.9498	0.9827	0.9384	0.6598	0.6563	0.5029	0.6783	0.7142	0.7545	0.8813	0.8408	0.6792	0.4346	0.6219	0.4775	0.7247	0.8801	0.9702
0.1101	17.0	6851	0.3954	0.7141	0.8198	0.9073	0.9718	0.7910	0.8139	0.6499	0.8004	0.8303	0.8739	0.9412	0.9122	0.8221	0.5549	0.7654	0.5929	0.8650	0.9493	0.9832	0.9385	0.6587	0.6547	0.5056	0.6806	0.7166	0.7550	0.8814	0.8395	0.6796	0.4414	0.6181	0.4756	0.7299	0.8802	0.9702
0.1095	18.0	7254	0.4032	0.7155	0.8201	0.9074	0.9678	0.8052	0.7992	0.6393	0.8075	0.8422	0.8699	0.9423	0.9094	0.8143	0.5625	0.7633	0.5948	0.8693	0.9512	0.9839	0.9373	0.6588	0.6628	0.5061	0.6808	0.7171	0.7553	0.8816	0.8412	0.6823	0.4440	0.6215	0.4785	0.7307	0.8796	0.9704
0.1066	19.0	7657	0.4130	0.7146	0.8192	0.9074	0.9693	0.7956	0.8023	0.6416	0.8124	0.8361	0.8702	0.9434	0.9117	0.8154	0.5534	0.7545	0.5971	0.8684	0.9509	0.9842	0.9380	0.6581	0.6587	0.5037	0.6791	0.7181	0.7561	0.8825	0.8409	0.6808	0.4397	0.6204	0.4779	0.7299	0.8798	0.9705
0.0992	20.0	8060	0.4199	0.7150	0.8193	0.9075	0.9704	0.7964	0.8037	0.6379	0.8041	0.8391	0.8711	0.9431	0.9116	0.8175	0.5566	0.7593	0.5995	0.8647	0.9506	0.9837	0.9380	0.6587	0.6591	0.5041	0.6800	0.7182	0.7558	0.8824	0.8412	0.6810	0.4420	0.6199	0.4787	0.7301	0.8798	0.9704

Framework versions

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