Create app.py

This is to use BLIP model to generate the caption from the Image uploaded through Gradio webUI, and playback speech by TTS at the same time.

app.py

@@ -0,0 +1,41 @@
+import gradio as gr
+from transformers import BlipForConditionalGeneration, BlipProcessor
+import torch
+import tempfile
+from gtts import gTTS
+
+# Load models
+device = "cpu"
+processor = BlipProcessor.from_pretrained("Salesforce/blip-image-captioning-large")
+model_image_captioning = BlipForConditionalGeneration.from_pretrained("Salesforce/blip-image-captioning-large").to(device)
+
+def generate_caption_tts(image):
+
+  inputs = processor(images=image, return_tensors="pt")
+  inputs["max_length"] = 20
+  inputs["num_beams"] = 5
+  outputs = model_image_captioning.generate(**inputs)
+
+  caption = processor.batch_decode(outputs, skip_special_tokens=True)[0]
+
+  speech = gTTS(caption, lang="en")
+  tmp_file = tempfile.mkstemp()[1]
+  speech.save(tmp_file)
+
+  return (caption, tmp_file)
+
+
+title = "Alec图像理解器"
+
+description = "Bootstrapping Language-Image Pre-training model演示：引导语言图像预训练以实现统一视觉语言理解和生成。 请上传您的图像"
+
+iface = gr.Interface(
+  fn=generate_caption_tts,
+  title=title,
+  description=description,
+  inputs=gr.inputs.Image(shape=(224,224)),
+  outputs=["text", "audio"]
+)
+
+
+iface.launch(share=True, debug=True)