nithin2002/imagecaptioning

1import torch 2import re 3import gradio as gr4from transformers import AutoTokenizer, ViTFeatureExtractor, VisionEncoderDecoderModel 5 6device='cpu'7encoder_checkpoint = "nlpconnect/vit-gpt2-image-captioning"8decoder_checkpoint = "nlpconnect/vit-gpt2-image-captioning"9model_checkpoint = "nlpconnect/vit-gpt2-image-captioning"10feature_extractor = ViTFeatureExtractor.from_pretrained(encoder_checkpoint)11tokenizer = AutoTokenizer.from_pretrained(decoder_checkpoint)12model = VisionEncoderDecoderModel.from_pretrained(model_checkpoint).to(device)13 14 15def predict(image,max_length=64, num_beams=4):16  image = image.convert('RGB')17  image = feature_extractor(image, return_tensors="pt").pixel_values.to(device)18  clean_text = lambda x: x.replace('<|endoftext|>','').split('\n')[0]19  caption_ids = model.generate(image, max_length = max_length)[0]20  caption_text = clean_text(tokenizer.decode(caption_ids))21  return caption_text 22 23 24 25input = gr.components.Image(label="Upload your Image", type = 'pil', optional=True)26output = gr.components.Textbox(type="text",label="Captions")27examples = [f"example{i}.jpg" for i in range(11,17)]28 29description="caption generation"30title = "Image Captioning using CNN and LSTM"31 32article = "Mini project B-12 "33 34interface = gr.Interface(35        fn=predict,36        inputs = input,37        theme="grass",38        outputs=output,39        examples = examples,40        title=title,41        description=description,42        article = article,43    )44interface.launch(debug=True)45