Open-source Image Classification Model exper_batch_16_e8 - Free Deployment, Image Classification Accuracy Reaches 91.29%

Exper Batch 16 E8

Developed by sudo-s

An image classification model fine-tuned on the sudo-s/herbier_mesuem1 dataset based on google/vit-base-patch16-224-in21k, achieving 91.29% accuracy

Image Classification

Transformers

Open Source License:Apache-2.0 #Plant Image Classification #High-Accuracy ViT #Small Batch Training

Downloads 30

Release Time : 6/26/2022

Model Overview

This is an image classification model based on the Vision Transformer architecture, specifically optimized for plant specimen images, suitable for scenarios such as digital classification of museum specimens.

Model Features

High Accuracy

Achieves 91.29% classification accuracy on the evaluation set

Efficient Fine-Tuning

Efficient fine-tuning based on a pre-trained ViT model, requiring only 8 epochs to achieve good results

Optimized Training

Uses mixed-precision training (Apex O1) and a linear learning rate scheduler to optimize the training process

Model Capabilities

Plant Specimen Image Classification

Museum Collection Digitization

Fine-Grained Image Recognition

Use Cases

Museum Digitization

Automatic Plant Specimen Classification

Automatically classify and label plant specimen images from museum collections

Accuracy reaches 91.29%

Biodiversity Research

Plant Species Identification

Assist researchers in quickly identifying and classifying plant samples

🚀 exper_batch_16_e8

This model is a fine - tuned version of [google/vit - base - patch16 - 224 - in21k](https://huggingface.co/google/vit - base - patch16 - 224 - in21k) on the sudo - s/herbier_mesuem1 dataset. It offers significant value in image - classification tasks by leveraging the pre - trained features of the base model and adapting them to the specific dataset.

🚀 Quick Start

This model is a fine - tuned version of [google/vit - base - patch16 - 224 - in21k](https://huggingface.co/google/vit - base - patch16 - 224 - in21k) on the sudo - s/herbier_mesuem1 dataset. It achieves the following results on the evaluation set:

Loss: 0.3951
Accuracy: 0.9129

📚 Documentation

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: 0.0002
train_batch_size: 16
eval_batch_size: 8
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon = 1e - 08
lr_scheduler_type: linear
num_epochs: 8
mixed_precision_training: Apex, opt level O1

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
3.8115	0.16	100	3.7948	0.1862
3.1194	0.31	200	3.0120	0.3281
2.3703	0.47	300	2.4791	0.4426
2.07	0.63	400	2.1720	0.5
1.6847	0.78	500	1.7291	0.5956
1.3821	0.94	600	1.4777	0.6299
0.9498	1.1	700	1.2935	0.6681
0.8741	1.25	800	1.1353	0.7051
0.8875	1.41	900	0.9951	0.7448
0.7233	1.56	1000	0.9265	0.7487
0.6696	1.72	1100	0.8660	0.7625
0.7364	1.88	1200	0.8710	0.7579
0.3933	2.03	1300	0.7162	0.8038
0.3443	2.19	1400	0.6305	0.8300
0.3376	2.35	1500	0.6273	0.8315
0.3071	2.5	1600	0.5988	0.8319
0.2863	2.66	1700	0.6731	0.8153
0.3017	2.82	1800	0.6042	0.8315
0.2382	2.97	1900	0.5118	0.8712
0.1578	3.13	2000	0.4917	0.8736
0.1794	3.29	2100	0.5302	0.8631
0.1093	3.44	2200	0.5035	0.8635
0.1076	3.6	2300	0.5186	0.8674
0.1219	3.76	2400	0.4723	0.8801
0.1017	3.91	2500	0.5132	0.8712
0.0351	4.07	2600	0.4709	0.8728
0.0295	4.23	2700	0.4674	0.8824
0.0416	4.38	2800	0.4836	0.8805
0.0386	4.54	2900	0.4663	0.8828
0.0392	4.69	3000	0.4003	0.8990
0.0383	4.85	3100	0.4187	0.8948
0.0624	5.01	3200	0.4460	0.8874
0.0188	5.16	3300	0.4169	0.9029
0.0174	5.32	3400	0.4098	0.8951
0.0257	5.48	3500	0.4289	0.8951
0.0123	5.63	3600	0.4295	0.9029
0.0052	5.79	3700	0.4395	0.8994
0.0081	5.95	3800	0.4217	0.9082
0.0032	6.1	3900	0.4216	0.9056
0.0033	6.26	4000	0.4113	0.9082
0.0024	6.42	4100	0.4060	0.9102
0.0022	6.57	4200	0.4067	0.9090
0.0031	6.73	4300	0.4005	0.9113
0.0021	6.89	4400	0.4008	0.9129
0.0021	7.04	4500	0.3967	0.9113
0.0043	7.2	4600	0.3960	0.9121
0.0022	7.36	4700	0.3962	0.9125
0.0021	7.51	4800	0.3992	0.9121
0.002	7.67	4900	0.3951	0.9129
0.0023	7.82	5000	0.3952	0.9125
0.0021	7.98	5100	0.3952	0.9129

Framework versions

Transformers 4.19.4
Pytorch 1.5.1
Datasets 2.3.2
Tokenizers 0.12.1

📄 License

This model is licensed under the Apache 2.0 license.

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