Upload app.py

app.py

@@ -1,53 +1,20 @@
 import gradio as gr
-from transformers import AutoProcessor, BlipForConditionalGeneration
 
-# from transformers import AutoProcessor, AutoTokenizer, AutoImageProcessor, AutoModelForCausalLM, BlipForConditionalGeneration, Blip2ForConditionalGeneration, VisionEncoderDecoderModel
-import torch
+import requests
+from PIL import Image
+from transformers import BlipProcessor, BlipForConditionalGeneration
 
-blip_processor_large = AutoProcessor.from_pretrained("umm-maybe/image-generator-identifier")
-blip_model_large = BlipForConditionalGeneration.from_pretrained("umm-maybe/image-generator-identifier")
+processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
+model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base")
 
-device = "cuda" if torch.cuda.is_available() else "cpu"
+def caption_image(raw_image):
+    inputs = processor(raw_image, return_tensors="pt")
+    out = model.generate(**inputs)
+    return processor.decode(out[0], skip_special_tokens=True)
 
-blip_model_large.to(device)
-
-def generate_caption(processor, model, image, tokenizer=None, use_float_16=False):
-    inputs = processor(images=image, return_tensors="pt").to(device)
-
-    if use_float_16:
-        inputs = inputs.to(torch.float16)
-    
-    generated_ids = model.generate(pixel_values=inputs.pixel_values, max_length=50)
-
-    if tokenizer is not None:
-        generated_caption = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
-    else:
-        generated_caption = processor.batch_decode(generated_ids, skip_special_tokens=True)[0]
-   
-    return generated_caption
-
-
-def generate_captions(image):
-
-    caption_blip_large = generate_caption(blip_processor_large, blip_model_large, image)
-
-    return caption_blip_large
-
-
-   
-examples = [["australia.jpg"], ["biden.png"], ["elon.jpg"], ["horns.jpg"], ["man.jpg"], ["nun.jpg"], ["painting.jpg"], ["pentagon.jpg"], ["pollock.jpg"], ["radcliffe.jpg"], ["split.jpg"], ["waves.jpg"], ["yeti.jpg"]]
 outputs = [
-           gr.outputs.Textbox(label="Caption including detected generator (if applicable)"),
+           gr.outputs.Textbox(label="Caption, including detected generator (if applicable)"),
           ] 
 
-title = "Generator Identification via Image Captioning"
-description = "Gradio Demo to illustrate the use of a fine-tuned BLIP image captioning to identify synthetic images.  To use it, simply upload your image and click 'submit', or click one of the examples to load them."
-
-interface = gr.Interface(fn=generate_captions, 
-                         inputs=gr.inputs.Image(type="pil"),
-                         outputs=outputs,
-                         examples=examples, 
-                         title=title,
-                         description=description,
-                         enable_queue=True)
-interface.launch()
+demo = gr.Interface(fn=caption_image, inputs="image", outputs=outputs)
+demo.launch()