pszemraj/beit-large-patch16-512-film-shot-classifier

beit-large-patch16-512: types of film shots

Model description

This model is a fine-tuned version of microsoft/beit-large-patch16-512 on the szymonrucinski/types-of-film-shots dataset. It achieves the following results on the evaluation set:

• Loss: 1.2335
• Accuracy: 0.6763

usage

from transformers import pipeline
from PIL import Image
import requests

pipe = pipeline(
    "image-classification",
    model="pszemraj/beit-large-patch16-512-film-shot-classifier",
)

url = "https://cdn-uploads.huggingface.co/production/uploads/60bccec062080d33f875cd0c/9YqYvv188ZccCMSzuv0KW.png"
image = Image.open(requests.get(url, stream=True).raw)
result = pipe(image)[0]
print(result)

try some of these:

class labels

The dataset contains the following labels:

"id2label": {
    "0": "ambiguous",
    "1": "closeUp",
    "2": "detail",
    "3": "extremeLongShot",
    "4": "fullShot",
    "5": "longShot",
    "6": "mediumCloseUp",
    "7": "mediumShot"
  },

as plaintext:

ambiguous, close up, detail, extreme long shot, full shot, long shot, medium close up, medium shot

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 2
• eval_batch_size: 4
• seed: 24414
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 6.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
1.0435	1.0	393	1.4799	0.4892
1.1554	2.0	786	1.4938	0.4892
1.5041	3.0	1179	2.1702	0.3597
1.0457	4.0	1572	1.5413	0.5683
0.3315	5.0	1965	1.0769	0.6978
0.2178	6.0	2358	1.2335	0.6763

Framework versions

• Transformers 4.38.0.dev0
• Pytorch 2.2.0+cu121
• Datasets 2.17.1
• Tokenizers 0.15.2