Upload 15 files

.gitattributes

@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+painting.png filter=lfs diff=lfs merge=lfs -text
+waves.png filter=lfs diff=lfs merge=lfs -text

app.py

@@ -1,20 +1,53 @@
 import gradio as gr
+from transformers import AutoProcessor, BlipForConditionalGeneration
 
-import requests
-from PIL import Image
-from transformers import BlipProcessor, BlipForConditionalGeneration
+# from transformers import AutoProcessor, AutoTokenizer, AutoImageProcessor, AutoModelForCausalLM, BlipForConditionalGeneration, Blip2ForConditionalGeneration, VisionEncoderDecoderModel
+import torch
 
-processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
-model = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base")
+blip_processor_large = AutoProcessor.from_pretrained("umm-maybe/image-generator-identifier")
+blip_model_large = BlipForConditionalGeneration.from_pretrained("umm-maybe/image-generator-identifier")
 
-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)
+device = "cuda" if torch.cuda.is_available() else "cpu"
 
+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)"),
           ] 
 
-demo = gr.Interface(fn=caption_image, inputs="image", outputs=outputs)
-demo.launch()
+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()

requirements.txt

@@ -1,5 +1,3 @@
+git+https://github.com/huggingface/transformers.git@main
 torch
-gradio
-requests
-pillow
-transformers
+torchvision