SRDdev/Image-Caption

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=3):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.inputs.Image(label="Upload any Image", type = 'pil', optional=True)26output = gr.outputs.Textbox(type="auto",label="Captions")27examples = [f"example{i}.jpg" for i in range(1,7)]28 29title = "Image Captioning "30description = "Made by : shreyasdixit.tech"31interface = gr.Interface(32            33        fn=predict,34        description=description,35        inputs = input,36        theme="grass",37        outputs=output,38        examples = examples,39        title=title,40    )41interface.launch(debug=True)