mit-b0-Image_segmentation_Dominoes_v2

This model is a fine-tuned version of nvidia/mit-b0.

It achieves the following results on the evaluation set:

• Loss: 0.1149
• Mean Iou: 0.9198
• Mean Accuracy: 0.9515
• Overall Accuracy: 0.9778
• Per Category Iou:
  • Segment 0: 0.974110559111975
  • Segment 1: 0.8655745252092782
• Per Category Accuracy
  • Segment 0: 0.9897833441005461
  • Segment 1: 0.913253525550903

Model description

For more information on how it was created, check out the following link: https://github.com/DunnBC22/Vision_Audio_and_Multimodal_Projects/blob/main/Computer%20Vision/Image%20Segmentation/Dominoes/Fine-Tuning%20-%20Dominoes%20-%20Image%20Segmentation%20with%20LoRA.ipynb

Intended uses & limitations

This model is intended to demonstrate my ability to solve a complex problem using technology.

Training and evaluation data

Dataset Source: https://huggingface.co/datasets/adelavega/dominoes_raw

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0005
• 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
• num_epochs: 25

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Per Category Iou Segment 0	Per Category Iou Segment 1	Per Category Accuracy Segment 0	Per Category Accuracy Segment 1
0.0461	1.0	86	0.1233	0.9150	0.9527	0.9762	0.9721967854031923	0.8578619172251059	0.9869082633464498	0.9184139264010376
0.0708	2.0	172	0.1366	0.9172	0.9490	0.9771	0.9732821853093164	0.8611008788165083	0.9898473600751747	0.9082362492748777
0.048	3.0	258	0.1260	0.9199	0.9534	0.9777	0.9740118174014271	0.8658241844233872	0.9888392553004053	0.9179240730467295
0.0535	4.0	344	0.1184	0.9200	0.9520	0.9778	0.974142444792198	0.8658711064023369	0.9896291184589182	0.9142864290038782
0.0185	5.0	430	0.1296	0.9182	0.9477	0.9775	0.9737715695013129	0.8627108292167807	0.9910418746696423	0.904378218719681
0.036	6.0	516	0.1410	0.9213	0.9538	0.9782	0.9745002408443008	0.8680673581922554	0.9892677512186527	0.9182967669045321
0.0376	7.0	602	0.1451	0.9206	0.9550	0.9779	0.9741455743906073	0.8669703237367214	0.9883004639689904	0.9216576612178001
0.0186	8.0	688	0.1380	0.9175	0.9496	0.9772	0.9733616852468584	0.8616466350192237	0.9897043519116697	0.9094762400541087
0.0162	9.0	774	0.1459	0.9218	0.9539	0.9783	0.9746840649852051	0.8688930149000804	0.989455276913138	0.9182917005479264
0.0169	10.0	860	0.1467	0.9191	0.9502	0.9776	0.9739086600912814	0.8642187978193332	0.9901195747929759	0.9102564589713776
0.0102	11.0	946	0.1549	0.9191	0.9524	0.9775	0.9737696499931041	0.8644247331609153	0.9889789745698009	0.915789237032027
0.0204	12.0	1032	0.1502	0.9215	0.9527	0.9783	0.974639596078376	0.8682964916021273	0.989902977623774	0.9155653673995151
0.0268	13.0	1118	0.1413	0.9194	0.9505	0.9777	0.9740020531855834	0.8647199376136	0.99011699066189	0.9107963425971664
0.0166	14.0	1204	0.1584	0.9173	0.9518	0.9770	0.9731154475737929	0.8614276032542578	0.9884142831972749	0.9152366875147241
0.0159	15.0	1290	0.1563	0.9170	0.9492	0.9770	0.9731832402253996	0.8607442858381036	0.9896456803899689	0.9087960816798012
0.0211	16.0	1376	0.1435	0.9150	0.9481	0.9764	0.9725201360275898	0.8574847000491036	0.989323310037	0.9068449010920532
0.0128	17.0	1462	0.1421	0.9212	0.9519	0.9782	0.9745789801464504	0.8677394402794754	0.9901920479238856	0.9136255861141298
0.0167	18.0	1548	0.1558	0.9217	0.9532	0.9783	0.9746811993626879	0.8686470009484697	0.9897428202266988	0.9166850322093621
0.0201	19.0	1634	0.1623	0.9156	0.9484	0.9766	0.9727184720007118	0.8584339325695252	0.9894484642039114	0.9072695251050635
0.0133	20.0	1720	0.1573	0.9189	0.9505	0.9776	0.9738320500157303	0.8640203613069115	0.9898665061373113	0.9112263496140702
0.012	21.0	1806	0.1631	0.9165	0.9472	0.9769	0.9731344243001482	0.8597866189796295	0.9904592118400188	0.9040137576913626
0.0148	22.0	1892	0.1629	0.9181	0.9507	0.9773	0.9735162429121835	0.8627239955489192	0.9894034768309156	0.9120129014770962
0.0137	23.0	1978	0.1701	0.9136	0.9484	0.9760	0.9719681843338751	0.8552607882028388	0.9885083690609032	0.908250815050119
0.0142	24.0	2064	0.1646	0.9146	0.9488	0.9763	0.9723134197764093	0.8568918401744342	0.9887405884771245	0.9089100747034281
0.0156	25.0	2150	0.1615	0.9144	0.9465	0.9763	0.9723929259786395	0.856345354289624	0.9898487696012216	0.9032139066422469

Framework versions

• Transformers 4.26.1
• Pytorch 2.0.1
• Datasets 2.13.1
• Tokenizers 0.13.3