git-base-pokemon Open-source Image Description Model - Accurately Generate Pokémon Image Description Content

Home

Git Base Pokemon

Developed by jihwaneom

An image caption generation model fine-tuned from microsoft/git-base, trained on Pokemon image dataset

Image-to-Text

Transformers

OtherOpen Source License:MIT #Image Caption Generation #Low Word Error Rate #Fine-tuned Model

Downloads 14

Release Time : 4/13/2023

Model Overview

This model is a vision-language model for generating image descriptions, specifically optimized for Pokemon images.

Model Features

Specialized for Pokemon Images

Description generation capability specifically optimized for Pokemon images

Low Word Error Rate

Achieved a word error rate of 1.9591 on the evaluation set

End-to-End Training

Trained end-to-end to directly learn the mapping from images to text

Model Capabilities

Image Caption Generation

Visual Content Understanding

Natural Language Generation

Use Cases

Game Content Generation

Automatic Pokedex Description

Automatically generate descriptive text for characters in Pokemon games

Can generate descriptions with relatively high accuracy

Creative Assistance

Fan Creation Support

Provide automatic descriptions for Pokemon fan works

🚀 git-base-pokemon

This model is a fine - tuned version of [microsoft/git - base](https://huggingface.co/microsoft/git - base) on the imagefolder dataset. It can achieve certain evaluation results, providing a solution for relevant image - related tasks.

🚀 Quick Start

This model is a fine - tuned version of [microsoft/git - base](https://huggingface.co/microsoft/git - base) on the imagefolder dataset. It achieves the following results on the evaluation set:

Loss: 0.0429
Wer Score: 1.9591

📚 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: 5e - 05
train_batch_size: 8
eval_batch_size: 8
seed: 42
gradient_accumulation_steps: 2
total_train_batch_size: 16
optimizer: Adam with betas=(0.9,0.999) and epsilon = 1e - 08
lr_scheduler_type: linear
num_epochs: 50
mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Score
7.3666	1.06	50	4.4430	21.5287
2.1581	2.13	100	0.2911	0.9783
0.0896	3.19	150	0.0328	0.3665
0.0269	4.26	200	0.0274	0.3487
0.0208	5.32	250	0.0284	0.4189
0.0168	6.38	300	0.0287	1.1673
0.0133	7.45	350	0.0296	6.0881
0.0106	8.51	400	0.0306	1.7969
0.0076	9.57	450	0.0322	7.1852
0.0053	10.64	500	0.0329	14.8889
0.0039	11.7	550	0.0338	12.2720
0.0027	12.77	600	0.0356	5.1533
0.0016	13.83	650	0.0371	8.4253
0.001	14.89	700	0.0379	6.7344
0.0006	15.96	750	0.0385	7.7586
0.0005	17.02	800	0.0392	9.0294
0.0004	18.09	850	0.0385	7.5083
0.0004	19.15	900	0.0394	5.1188
0.0004	20.21	950	0.0397	5.0600
0.0004	21.28	1000	0.0399	4.4125
0.0003	22.34	1050	0.0405	3.7803
0.0003	23.4	1100	0.0406	3.3397
0.0003	24.47	1150	0.0408	3.3218
0.0003	25.53	1200	0.0411	2.8212
0.0003	26.6	1250	0.0411	2.7165
0.0003	27.66	1300	0.0414	2.7625
0.0003	28.72	1350	0.0416	2.4330
0.0003	29.79	1400	0.0416	2.2350
0.0003	30.85	1450	0.0419	2.1699
0.0003	31.91	1500	0.0421	2.0026
0.0003	32.98	1550	0.0420	2.1609
0.0003	34.04	1600	0.0421	2.0307
0.0003	35.11	1650	0.0422	1.9668
0.0003	36.17	1700	0.0423	1.9387
0.0003	37.23	1750	0.0425	1.9464
0.0003	38.3	1800	0.0427	1.8761
0.0003	39.36	1850	0.0427	1.8940
0.0003	40.43	1900	0.0428	1.9068
0.0003	41.49	1950	0.0428	1.8774
0.0003	42.55	2000	0.0429	1.8352
0.0002	43.62	2050	0.0428	2.0907
0.0002	44.68	2100	0.0429	2.0319
0.0002	45.74	2150	0.0429	2.0179
0.0002	46.81	2200	0.0429	1.9706
0.0002	47.87	2250	0.0429	1.9604
0.0002	48.94	2300	0.0429	1.9540
0.0002	50.0	2350	0.0429	1.9591

Framework versions

Transformers 4.28.0
Pytorch 2.0.0+cu118
Datasets 2.11.0
Tokenizers 0.13.3

📄 License

This project is licensed under the MIT 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