rvv-karma's picture
Update README.md
fb62bce
metadata
license: apache-2.0
base_model: google/vit-base-patch16-224
tags:
  - generated_from_trainer
metrics:
  - accuracy
model-index:
  - name: Human-Action-Recognition-VIT-Base-patch16-224
    results: []
datasets:
  - Bingsu/Human_Action_Recognition
language:
  - en
pipeline_tag: image-classification

Human-Action-Recognition-VIT-Base-patch16-224

This model is a fine-tuned version of google/vit-base-patch16-224 on Bingsu/Human_Action_Recognition dataset.
It achieves the following results on the evaluation set:

  • Loss: 0.4005
  • Accuracy: 0.8786

Model description

The Vision Transformer (ViT) is a transformer encoder model (BERT-like) pretrained on a large collection of images in a supervised fashion, namely ImageNet-21k, at a resolution of 224x224 pixels. Next, the model was fine-tuned on ImageNet (also referred to as ILSVRC2012), a dataset comprising 1 million images and 1,000 classes, also at resolution 224x224.

Images are presented to the model as a sequence of fixed-size patches (resolution 16x16), which are linearly embedded. One also adds a [CLS] token to the beginning of a sequence to use it for classification tasks. One also adds absolute position embeddings before feeding the sequence to the layers of the Transformer encoder.

By pre-training the model, it learns an inner representation of images that can then be used to extract features useful for downstream tasks: if you have a dataset of labeled images for instance, you can train a standard classifier by placing a linear layer on top of the pre-trained encoder. One typically places a linear layer on top of the [CLS] token, as the last hidden state of this token can be seen as a representation of an entire image.

Intended uses & limitations

You can use the model for image classification.

How to use

Here is how to use this model to classify an image of the human action into one of the following categories:
calling, clapping, cycling, dancing, drinking, eating, fighting, hugging, laughing, listening_to_music, running, sitting, sleeping, texting, using_laptop

from transformers import pipeline
from PIL import Image
import requests

pipe = pipeline("image-classification", "rvv-karma/Human-Action-Recognition-VIT-Base-patch16-224")
url = "https://images.pexels.com/photos/175658/pexels-photo-175658.jpeg?auto=compress&cs=tinysrgb&dpr=1&w=500"
image = Image.open(requests.get(url, stream=True).raw)
pipe(image)

# Output:
# [{'score': 0.9918079972267151, 'label': 'dancing'},
#  {'score': 0.00207977625541389, 'label': 'clapping'},
#  {'score': 0.0015223610680550337, 'label': 'running'},
#  {'score': 0.0009153694845736027, 'label': 'fighting'},
#  {'score': 0.0006987180095165968, 'label': 'sitting'}]

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: 64
  • eval_batch_size: 64
  • seed: 42
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 256
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_ratio: 0.1
  • num_epochs: 20

Training results

Training Loss Epoch Step Validation Loss Accuracy
2.6396 0.99 39 2.0436 0.4425
1.4579 2.0 79 0.7553 0.7917
0.8342 2.99 118 0.5296 0.8417
0.6649 4.0 158 0.4978 0.8496
0.6137 4.99 197 0.4460 0.8595
0.5374 6.0 237 0.4356 0.8627
0.514 6.99 276 0.4349 0.8615
0.475 8.0 316 0.4005 0.8786
0.4663 8.99 355 0.4164 0.8659
0.4178 10.0 395 0.4128 0.8738
0.4226 10.99 434 0.4115 0.8690
0.3896 12.0 474 0.4112 0.875
0.3866 12.99 513 0.4072 0.8714
0.3632 14.0 553 0.4106 0.8718
0.3596 14.99 592 0.4043 0.8714
0.3421 16.0 632 0.4128 0.8675
0.344 16.99 671 0.4181 0.8643
0.3447 18.0 711 0.4128 0.8687
0.3407 18.99 750 0.4097 0.8714
0.3267 19.75 780 0.4097 0.8683

Framework versions

  • Transformers 4.35.2
  • Pytorch 2.1.0+cu118
  • Datasets 2.15.0
  • Tokenizers 0.15.0

Fine-tuning script

Google Colaboratory Notebook