Model Details
Model type: Vision Transformer (ViT) for Image Classification
Finetuned from model : google/vit-base-patch16-384
Uses
Image classification based on facial features from the dataset.Link:https://www.kaggle.com/datasets/bhaveshmittal/celebrity-face-recognition-dataset
Downstream Use
Fine-tuning for other image classification tasks.
Transfer learning for related vision tasks.
Out-of-Scope Use
Tasks unrelated to image classification.
Sensitive applications without proper evaluation of biases and limitations.
Bias, Risks, and Limitations
Potential biases in the training dataset affecting model predictions.
Limitations in generalizability to different populations or image conditions not represented in the training data.
Risks associated with misclassification in critical applications.
Recommendations
Users (both direct and downstream) should be made aware of the risks, biases, and limitations of the model. It's recommended to evaluate the model's performance on the specific data before deploying it in a production environment
How to Get Started with the Model
Use the code below to get started with the model.
import torch
from transformers import ViTForImageClassification, ViTImageProcessor
model = ViTForImageClassification.from_pretrained("Ganesh-KSV/face-recognition-version1")
processor = ViTImageProcessor.from_pretrained("Ganesh-KSV/face-recognition-version1")
def predict(image):
inputs = processor(images=image, return_tensors="pt")
outputs = model(**inputs)
logits = outputs.logits
predictions = torch.argmax(logits, dim=-1)
return predictions
Training Details
Training Data
Training Procedure:
Preprocessing :
Images were resized, augmented (rotation, color jitter, etc.), and normalized.
Training Hyperparameters:
Optimizer: Adam with learning rate 2e-5 and weight decay 1e-2
Scheduler: StepLR with step size 2 and gamma 0.5
Loss Function: CrossEntropyLoss
Epochs: 40
Batch Size: 4
Evaluation
Testing Data, Factors & Metrics
Testing Data
Validation split of the VGGFace dataset.
Factors
Performance evaluated based on loss and accuracy on the validation set.
Metrics
Loss and accuracy metrics for each epoch.
Results
Training and validation loss and accuracy plotted for 40 epochs.
Confusion matrix generated for the final validation results.
Summary
Model Examination
Model performance examined through loss, accuracy plots, and confusion matrix.
Glossary
ViT: Vision Transformer
CrossEntropyLoss: A loss function used for classification tasks.
Adam: An optimization algorithm.
StepLR: Learning rate scheduler that decays the learning rate by a factor every few epochs.
- Downloads last month
- 244