Ramos-Ramos/emb-gam-dino

Model description

This is a LogisticRegressionCV model trained on averages of patch embeddings from the Imagenette dataset. This forms the GAM of an Emb-GAM extended to images. Patch embeddings are meant to be extracted with the facebook/dino-vitb16 DINO checkpoint.

Intended uses & limitations

This model is not intended to be used in production.

Training Procedure

Hyperparameters

The model is trained with below hyperparameters.

Click to expand

Hyperparameter	Value
Cs	10
class_weight
cv	StratifiedKFold(n_splits=5, random_state=1, shuffle=True)
dual	False
fit_intercept	True
intercept_scaling	1.0
l1_ratios
max_iter	100
multi_class	auto
n_jobs
penalty	l2
random_state	1
refit	False
scoring
solver	lbfgs
tol	0.0001
verbose	0

Model Plot

The model plot is below.

LogisticRegressionCV(cv=StratifiedKFold(n_splits=5, random_state=1, shuffle=True),random_state=1, refit=False)

Please rerun this cell to show the HTML repr or trust the notebook.

Evaluation Results

You can find the details about evaluation process and the evaluation results.

Metric	Value
accuracy	0.97707
f1 score	0.97707

How to Get Started with the Model

Use the code below to get started with the model.

Click to expand

from PIL import Image
from skops import hub_utils
import torch
from transformers import AutoFeatureExtractor, AutoModel
import pickle
import os

# load embedding model
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
feature_extractor = AutoFeatureExtractor.from_pretrained('facebook/dino-vitb16')
model = AutoModel.from_pretrained('facebook/dino-vitb16').eval().to(device)

# load logistic regression
os.mkdir('emb-gam-dino')
hub_utils.download(repo_id='Ramos-Ramos/emb-gam-dino', dst='emb-gam-dino')

with open('emb-gam-dino/model.pkl', 'rb') as file: 
  logistic_regression = pickle.load(file)
    
# load image
img = Image.open('examples/english_springer.png')

# preprocess image
inputs = {k: v.to(device) for k, v in feature_extractor(img, return_tensors='pt').items()}

# extract patch embeddings
with torch.no_grad():
  patch_embeddings = model(**inputs).last_hidden_state[0, 1:].cpu()

# classify
pred = logistic_regression.predict(patch_embeddings.sum(dim=0, keepdim=True))

# get patch contributions
patch_contributions = logistic_regression.coef_ @ patch_embeddings.T.numpy()

Model Card Authors

This model card is written by following authors:

Patrick Ramos

Model Card Contact

You can contact the model card authors through following channels: [More Information Needed]

Citation

BibTeX:

@article{singh2022emb,
  title={Emb-GAM: an Interpretable and Efficient Predictor using Pre-trained Language Models},
  author={Singh, Chandan and Gao, Jianfeng},
  journal={arXiv preprint arXiv:2209.11799},
  year={2022}
}

Additional Content

confusion_matrix

Demo

Check out our HuggingFace Space here! It does Imagenette classification and visualizes patch contributions per label.