Update app.py

app.py

@@ -9,14 +9,17 @@ torch.hub.download_url_to_file('http://images.cocodataset.org/val2017/0000000397
 torch.hub.download_url_to_file('https://huggingface.co/datasets/nielsr/textcaps-sample/resolve/main/stop_sign.png', 'stop_sign.png')
 torch.hub.download_url_to_file('https://cdn.openai.com/dall-e-2/demos/text2im/astronaut/horse/photo/0.jpg', 'astronaut.jpg')
 
-git_processor_base = AutoProcessor.from_pretrained("microsoft/git-base-coco")
-git_model_base = AutoModelForCausalLM.from_pretrained("microsoft/git-base-coco")
+# git_processor_base = AutoProcessor.from_pretrained("microsoft/git-base-coco")
+# git_model_base = AutoModelForCausalLM.from_pretrained("microsoft/git-base-coco")
 
-git_processor_large = AutoProcessor.from_pretrained("microsoft/git-large-coco")
-git_model_large = AutoModelForCausalLM.from_pretrained("microsoft/git-large-coco")
+git_processor_large_coco = AutoProcessor.from_pretrained("microsoft/git-large-coco")
+git_model_large_coco = AutoModelForCausalLM.from_pretrained("microsoft/git-large-coco")
 
-blip_processor_base = AutoProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
-blip_model_base = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base")
+git_processor_large_textcaps = AutoProcessor.from_pretrained("microsoft/git-large-r-textcaps")
+git_model_large_textcaps = AutoModelForCausalLM.from_pretrained("microsoft/git-large-r-textcaps")
+
+# blip_processor_base = AutoProcessor.from_pretrained("Salesforce/blip-image-captioning-base")
+# blip_model_base = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-base")
 
 blip_processor_large = AutoProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
 blip_model_large = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large")
@@ -32,9 +35,10 @@ coca_model, _, coca_transform = open_clip.create_model_and_transforms(
 
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
-git_model_base.to(device)
-blip_model_base.to(device)
-git_model_large.to(device)
+# git_model_base.to(device)
+# blip_model_base.to(device)
+git_model_large_coco.to(device)
+git_model_large_textcaps.to(device)
 blip_model_large.to(device)
 vitgpt_model.to(device)
 coca_model.to(device)
@@ -60,11 +64,13 @@ def generate_caption_coca(model, transform, image):
 
 
 def generate_captions(image):
-    caption_git_base = generate_caption(git_processor_base, git_model_base, image)
+    # caption_git_base = generate_caption(git_processor_base, git_model_base, image)
+
+    caption_git_large_coco = generate_caption(git_processor_large_coco, git_model_large_coco, image)
 
-    caption_git_large = generate_caption(git_processor_large, git_model_large, image)
+    caption_git_large_textcaps = generate_caption(git_processor_large_textcaps, git_model_large_textcaps, image)
 
-    caption_blip_base = generate_caption(blip_processor_base, blip_model_base, image)
+    # caption_blip_base = generate_caption(blip_processor_base, blip_model_base, image)
 
     caption_blip_large = generate_caption(blip_processor_large, blip_model_large, image)
 
@@ -72,11 +78,11 @@ def generate_captions(image):
 
     caption_coca = generate_caption_coca(coca_model, coca_transform, image)
 
-    return caption_git_base, caption_git_large, caption_blip_base, caption_blip_large, caption_vitgpt, caption_coca
+    return caption_git_large_coco, caption_git_large_textcaps, caption_blip_large, caption_vitgpt, caption_coca
 
    
 examples = [["cats.jpg"], ["stop_sign.png"], ["astronaut.jpg"]]
-outputs = [gr.outputs.Textbox(label="Caption generated by GIT-base"), gr.outputs.Textbox(label="Caption generated by GIT-large"), gr.outputs.Textbox(label="Caption generated by BLIP-base"), gr.outputs.Textbox(label="Caption generated by BLIP-large"), gr.outputs.Textbox(label="Caption generated by ViT+GPT-2"), gr.outputs.Textbox(label="Caption generated by CoCa")] 
+outputs = [gr.outputs.Textbox(label="Caption generated by GIT-base fine-tuned on COCO"), gr.outputs.Textbox(label="Caption generated by GIT-large fine-tuned on TextCaps"), gr.outputs.Textbox(label="Caption generated by BLIP-large"), gr.outputs.Textbox(label="Caption generated by ViT+GPT-2"), gr.outputs.Textbox(label="Caption generated by CoCa")] 
 
 title = "Interactive demo: comparing image captioning models"
 description = "Gradio Demo to compare GIT, BLIP, ViT+GPT2 and CoCa, 4 state-of-the-art vision+language models. To use it, simply upload your image and click 'submit', or click one of the examples to load them. Read more at the links below."