added 4 more models

app.py

@@ -1,5 +1,5 @@
 import gradio as gr
-from transformers import AutoProcessor, BlipForConditionalGeneration, AutoModelForCausalLM
+from transformers import AutoProcessor, BlipForConditionalGeneration, AutoModelForCausalLM, AutoImageProcessor, VisionEncoderDecoderModel, AutoTokenizer
 
 # from transformers import AutoProcessor, AutoTokenizer, AutoImageProcessor, AutoModelForCausalLM, BlipForConditionalGeneration, Blip2ForConditionalGeneration, VisionEncoderDecoderModel
 import torch
@@ -14,17 +14,17 @@ torch.hub.download_url_to_file('https://cdn.openai.com/dall-e-2/demos/text2im/as
 git_processor_base = AutoProcessor.from_pretrained("microsoft/git-base-coco")
 git_model_base = AutoModelForCausalLM.from_pretrained("microsoft/git-base-coco")
 
-# git_processor_large_coco = AutoProcessor.from_pretrained("microsoft/git-large-coco")
-# git_model_large_coco = 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")
 
-# git_processor_large_textcaps = AutoProcessor.from_pretrained("microsoft/git-large-r-textcaps")
-# git_model_large_textcaps = AutoModelForCausalLM.from_pretrained("microsoft/git-large-r-textcaps")
+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")
+blip_processor_large = AutoProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
+blip_model_large = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large")
 
 # blip2_processor = AutoProcessor.from_pretrained("Salesforce/blip2-opt-2.7b")
 # blip2_model = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-opt-2.7b", torch_dtype=torch.float16)
@@ -32,9 +32,9 @@ blip_model_base = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-
 # blip2_processor_8_bit = AutoProcessor.from_pretrained("Salesforce/blip2-opt-6.7b")
 # blip2_model_8_bit = Blip2ForConditionalGeneration.from_pretrained("Salesforce/blip2-opt-6.7b", device_map="auto", load_in_8bit=True)
 
-# vitgpt_processor = AutoImageProcessor.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
-# vitgpt_model = VisionEncoderDecoderModel.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
-# vitgpt_tokenizer = AutoTokenizer.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
+vitgpt_processor = AutoImageProcessor.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
+vitgpt_model = VisionEncoderDecoderModel.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
+vitgpt_tokenizer = AutoTokenizer.from_pretrained("nlpconnect/vit-gpt2-image-captioning")
 
 # coca_model, _, coca_transform = open_clip.create_model_and_transforms(
 #   model_name="coca_ViT-L-14",
@@ -78,15 +78,15 @@ 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_large_coco = generate_caption(git_processor_large_coco, git_model_large_coco, image)
+    caption_git_large_coco = generate_caption(git_processor_large_coco, git_model_large_coco, image)
 
-    # caption_git_large_textcaps = generate_caption(git_processor_large_textcaps, git_model_large_textcaps, 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_large = generate_caption(blip_processor_large, blip_model_large, image)
+    caption_blip_large = generate_caption(blip_processor_large, blip_model_large, image)
 
-    # caption_vitgpt = generate_caption(vitgpt_processor, vitgpt_model, image, vitgpt_tokenizer)
+    caption_vitgpt = generate_caption(vitgpt_processor, vitgpt_model, image, vitgpt_tokenizer)
 
     # caption_coca = generate_caption_coca(coca_model, coca_transform, image)
 
@@ -95,13 +95,19 @@ def generate_captions(image):
     # caption_blip2_8_bit = generate_caption(blip2_processor_8_bit, blip2_model_8_bit, image, use_float_16=True).strip()
 
     # return caption_git_large_coco, caption_git_large_textcaps, caption_blip_large, caption_coca, caption_blip2_8_bit
-    return caption_git_base, caption_blip_base
+    return caption_git_base, caption_git_large_coco, caption_git_large_textcaps, caption_blip_base, caption_blip_large, caption_vitgpt
 
 
    
 examples = [["cats.jpg"], ["stop_sign.png"], ["astronaut.jpg"]]
 # outputs = [gr.outputs.Textbox(label="Caption generated by GIT-large 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 CoCa"), gr.outputs.Textbox(label="Caption generated by BLIP-2 OPT 6.7b")] 
-outputs = [gr.outputs.Textbox(label="Caption generated by GIT-base fine-tuned on COCO"), gr.outputs.Textbox(label="Caption generated by BLIP-base")] 
+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 COCO"),
+           gr.outputs.Textbox(label="Caption generated by GIT-large fine-tuned on TextCaps"),
+           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 vitgpt")
+          ] 
 
 title = "Interactive demo: comparing image captioning models"
 description = "Gradio Demo to compare GIT, BLIP, CoCa, and BLIP-2, 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."