Open-source SegFormer-B0 Human Parsing Model - Free Deployment for Precise Identification of Human Body Parts in Images

Segformer B0 Finetuned Human Parsing

Developed by matei-dorian

A human parsing model fine-tuned based on SegFormer-B0 architecture for segmenting and recognizing human body parts in images

Image Segmentation

Transformers

Open Source License:Other #Human body part segmentation #Fine-grained semantic segmentation #Clothing recognition

Downloads 16

Release Time : 5/2/2023

Model Overview

This model is a fine-tuned version of the MIT-B0 architecture for human parsing tasks, capable of identifying and segmenting different human body parts (such as hair, tops, pants, etc.) in images

Model Features

Lightweight Architecture

Based on SegFormer-B0 architecture, maintains good performance while having a relatively small model size

Multi-category Segmentation

Capable of recognizing and segmenting 20+ human body parts and clothing categories

Transfer Learning

Fine-tuned from pre-trained MIT-B0 model, adapted for specific human parsing tasks

Model Capabilities

Image segmentation

Human body part recognition

Clothing classification

Semantic segmentation

Use Cases

Fashion & Retail

Virtual Try-on

Identify user's clothing for virtual garment overlay

Fashion Analysis

Analyze clothing combinations and fashion trends in images

Human-Computer Interaction

Augmented Reality Applications

Accurately recognize human body parts for more natural AR interactions

🚀 segformer-b0-finetuned-human-parsing

This model is a fine - tuned version of [nvidia/mit - b0](https://huggingface.co/nvidia/mit - b0). It is designed for image segmentation tasks, specifically human parsing. The model achieves certain performance metrics on the evaluation set, providing valuable insights into its effectiveness.

🚀 Quick Start

This model is a fine - tuned version of [nvidia/mit - b0](https://huggingface.co/nvidia/mit - b0) on the None dataset. It achieves the following results on the evaluation set:

Loss: 1.9476
Mean Iou: 0.0726
Mean Accuracy: 0.1221
Overall Accuracy: 0.3575
Accuracy Background: nan
Accuracy Hat: 0.0048
Accuracy Hair: 0.4813
Accuracy Sunglasses: 0.0
Accuracy Upper - clothes: 0.9405
Accuracy Skirt: 0.0000
Accuracy Pants: 0.0631
Accuracy Dress: 0.1031
Accuracy Belt: 0.0
Accuracy Left - shoe: 0.0011
Accuracy Right - shoe: 0.0010
Accuracy Face: 0.4406
Accuracy Left - leg: 0.0291
Accuracy Right - leg: 0.0
Accuracy Left - arm: 0.0
Accuracy Right - arm: 0.0001
Accuracy Bag: 0.0114
Accuracy Scarf: 0.0
Iou Background: 0.0
Iou Hat: 0.0043
Iou Hair: 0.4221
Iou Sunglasses: 0.0
Iou Upper - clothes: 0.3239
Iou Skirt: 0.0000
Iou Pants: 0.0559
Iou Dress: 0.0728
Iou Belt: 0.0
Iou Left - shoe: 0.0011
Iou Right - shoe: 0.0009
Iou Face: 0.3872
Iou Left - leg: 0.0271
Iou Right - leg: 0.0
Iou Left - arm: 0.0
Iou Right - arm: 0.0001
Iou Bag: 0.0106
Iou Scarf: 0.0

📚 Documentation

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 6e - 05
train_batch_size: 2
eval_batch_size: 2
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon = 1e - 08
lr_scheduler_type: linear
num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Background	Accuracy Hat	Accuracy Hair	Accuracy Sunglasses	Accuracy Upper - clothes	Accuracy Skirt	Accuracy Pants	Accuracy Dress	Accuracy Left - shoe	Accuracy Right - shoe	Accuracy Face	Accuracy Left - leg	Accuracy Right - leg	Accuracy Left - arm	Accuracy Right - arm	Accuracy Bag	Iou Hat	Iou Hair	Iou Sunglasses	Iou Upper - clothes	Iou Skirt	Iou Pants	Iou Dress	Iou Left - shoe	Iou Right - shoe	Iou Face	Iou Left - leg	Iou Right - leg	Iou Left - arm	Iou Right - arm	Iou Bag
2.5768	0.4	20	2.7812	0.0726	0.1332	0.2876	nan	0.0178	0.3204	0.0004	0.5548	0.0004	0.2555	0.2373	0.0103	0.0003	0.5637	0.0287	0.0302	0.0001	0.0008	0.2435	0.0166	0.2759	0.0001	0.2781	0.0004	0.1710	0.1295	0.0098	0.0003	0.3251	0.0260	0.0248	0.0001	0.0007	0.0491
2.2093	0.8	40	2.5166	0.0563	0.1052	0.3288	nan	0.0	0.1994	0.0	0.9447	0.0015	0.0435	0.1164	0.0008	0.0000	0.4655	0.0007	0.0003	0.0	0.0	0.0153	0.0	0.1946	0.0	0.3037	0.0015	0.0417	0.0842	0.0008	0.0000	0.3726	0.0007	0.0003	0.0	0.0	0.0124
1.8804	1.2	60	2.0209	0.0632	0.1110	0.3374	nan	0.0087	0.3724	0.0	0.9475	0.0014	0.0162	0.0528	0.0001	0.0008	0.4257	0.0561	0.0001	0.0	0.0	0.0055	0.0077	0.3472	0.0	0.3086	0.0014	0.0156	0.0403	0.0001	0.0008	0.3597	0.0515	0.0001	0.0	0.0	0.0052
1.8776	1.6	80	2.0016	0.0665	0.1154	0.3454	nan	0.0056	0.4172	0.0	0.9412	0.0000	0.0490	0.0697	0.0002	0.0006	0.4349	0.0329	0.0000	0.0	0.0000	0.0100	0.0048	0.3791	0.0	0.3138	0.0000	0.0438	0.0542	0.0002	0.0006	0.3608	0.0304	0.0000	0.0	0.0000	0.0093
1.8471	2.0	100	1.9476	0.0726	0.1221	0.3575	nan	0.0048	0.4813	0.0	0.9405	0.0000	0.0631	0.1031	0.0011	0.0010	0.4406	0.0291	0.0	0.0	0.0001	0.0114	0.0043	0.4221	0.0	0.3239	0.0000	0.0559	0.0728	0.0011	0.0009	0.3872	0.0271	0.0	0.0	0.0001	0.0106

Framework versions

Transformers 4.28.1
Pytorch 2.0.0+cu118
Datasets 2.12.0
Tokenizers 0.13.3

📄 License

License: other

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