finetune-instance-segmentation-flowchartseg-mask2former_20epochs_a6000

This model is a fine-tuned version of facebook/mask2former-swin-tiny-coco-instance on the MananSuri27/flowchartseg dataset. It achieves the following results on the evaluation set:

• Loss: 6.9523
• Map: 0.7683
• Map 50: 0.9698
• Map 75: 0.9012
• Map Small: 0.6166
• Map Medium: 0.8012
• Map Large: 0.9911
• Mar 1: 0.0414
• Mar 10: 0.3998
• Mar 100: 0.8409
• Mar Small: 0.6954
• Mar Medium: 0.8711
• Mar Large: 0.9938
• Map Per Class: 0.7683
• Mar 100 Per Class: 0.8409
• Classes: 0

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: 1e-05
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• num_epochs: 20.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Map	Map 50	Map 75	Map Small	Map Medium	Map Large	Mar 1	Mar 10	Mar 100	Mar Small	Mar Medium	Mar Large	Map Per Class	Mar 100 Per Class	Classes
21.8993	1.0	75	12.4440	0.6744	0.9348	0.7871	0.4962	0.7115	0.9595	0.0394	0.3779	0.7627	0.5896	0.7977	0.9741	0.6744	0.7627	0
10.836	2.0	150	10.1046	0.7114	0.9553	0.8305	0.5421	0.7495	0.9806	0.0399	0.3853	0.7924	0.6239	0.8267	0.9864	0.7114	0.7924	0
9.422	3.0	225	9.3849	0.7277	0.9599	0.8501	0.5591	0.7674	0.9829	0.0398	0.3882	0.8057	0.6454	0.8385	0.9852	0.7277	0.8057	0
8.7626	4.0	300	8.7372	0.7337	0.9632	0.8564	0.5685	0.7713	0.9825	0.0401	0.3902	0.8122	0.6565	0.8441	0.9889	0.7337	0.8122	0
8.2403	5.0	375	8.3938	0.7416	0.9673	0.8706	0.5659	0.7812	0.991	0.0403	0.3914	0.8177	0.657	0.8508	0.9926	0.7416	0.8177	0
7.8836	6.0	450	7.9279	0.7504	0.9692	0.8794	0.5853	0.7878	0.9901	0.0412	0.3964	0.824	0.6685	0.8561	0.9914	0.7504	0.824	0
7.6029	7.0	525	7.8302	0.7562	0.9707	0.8933	0.5965	0.7925	0.9909	0.0415	0.3969	0.8294	0.68	0.8603	0.9926	0.7562	0.8294	0
7.3823	8.0	600	7.5476	0.756	0.9689	0.8892	0.5973	0.792	0.9908	0.0412	0.3982	0.8315	0.6809	0.8626	0.9926	0.756	0.8315	0
7.2456	9.0	675	7.4617	0.7611	0.969	0.8979	0.6055	0.795	0.9902	0.0411	0.3995	0.8338	0.6865	0.8642	0.9926	0.7611	0.8338	0
7.068	10.0	750	7.3238	0.7627	0.9688	0.8988	0.6028	0.7964	0.9901	0.0413	0.3995	0.8347	0.685	0.8656	0.9938	0.7627	0.8347	0
7.0105	11.0	825	7.2475	0.7655	0.9703	0.8992	0.6099	0.7982	0.991	0.0409	0.4007	0.8372	0.6893	0.8679	0.9938	0.7655	0.8372	0
6.9651	12.0	900	7.2345	0.7679	0.9709	0.9045	0.6147	0.8008	0.9899	0.0412	0.3999	0.8393	0.6944	0.8692	0.9938	0.7679	0.8393	0
6.8849	13.0	975	7.0420	0.764	0.9698	0.8996	0.6091	0.7977	0.9899	0.0405	0.3996	0.8377	0.692	0.8677	0.9938	0.764	0.8377	0
6.8559	14.0	1050	7.0714	0.7643	0.9693	0.8986	0.6084	0.7991	0.9899	0.0413	0.3994	0.8387	0.6913	0.8693	0.9938	0.7643	0.8387	0
6.7735	15.0	1125	6.9974	0.7671	0.9702	0.9007	0.6119	0.8014	0.9908	0.0413	0.3998	0.8395	0.6902	0.8706	0.9938	0.7671	0.8395	0
6.7596	16.0	1200	6.9287	0.7675	0.9702	0.9041	0.614	0.8007	0.991	0.0406	0.399	0.8393	0.6931	0.8696	0.9938	0.7675	0.8393	0
6.7191	17.0	1275	7.0029	0.7678	0.9701	0.9039	0.61	0.8016	0.991	0.0407	0.3991	0.8407	0.6943	0.8711	0.9938	0.7678	0.8407	0
6.6811	18.0	1350	6.9408	0.7678	0.97	0.901	0.6121	0.8014	0.991	0.041	0.4008	0.84	0.6928	0.8705	0.9938	0.7678	0.84	0
6.6732	19.0	1425	6.9605	0.7679	0.9696	0.9012	0.6142	0.801	0.9911	0.0413	0.4005	0.8405	0.6935	0.871	0.9938	0.7679	0.8405	0
6.7621	19.7407	1480	6.9523	0.7683	0.9698	0.9012	0.6166	0.8012	0.9911	0.0414	0.3998	0.8409	0.6954	0.8711	0.9938	0.7683	0.8409	0

Framework versions

• Transformers 4.49.0.dev0
• Pytorch 2.4.0+cu121
• Datasets 2.21.0
• Tokenizers 0.21.0