Update app.py

app.py

@@ -1,25 +1,24 @@
 import streamlit as st
 from PIL import Image
 import torch
-from transformers import pipeline
-import numpy as np
+from diffusers import DiffusionPipeline
 import io
+import numpy as np
 
 # إعداد الصفحة
-st.set_page_config(page_title="Image to Video with Editing", page_icon="🎥")
+st.set_page_config(page_title="InstantMesh - 3D Mesh Generation", page_icon="🖼️")
 
 # عنوان الصفحة
-st.title("Stable Video Diffusion - Image to Video")
+st.title("InstantMesh - 3D Mesh Generation from Image")
 
 # تعليمات
-st.write("Upload an image to generate a video. You can also adjust settings for video generation.")
+st.write("Upload an image to generate a 3D mesh. InstantMesh creates detailed 3D models within seconds.")
 
 # تحميل النموذج
 @st.cache_resource
 def load_model():
-    model_id = "stabilityai/stable-video-diffusion-img2vid-xt"
-    # تأكد من استخدام الكلاس الصحيح من مكتبة transformers إذا كان متاحًا
-    pipe = pipeline("image-to-video", model=model_id, torch_dtype=torch.float16).to("cuda")
+    model_id = "TencentARC/InstantMesh"
+    pipe = DiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16).to("cuda")
     return pipe
 
 pipe = load_model()
@@ -27,32 +26,33 @@ pipe = load_model()
 # إدخال المستخدم
 uploaded_image = st.file_uploader("Choose an image...", type=["jpg", "jpeg", "png"])
 
-# إعداد الخيارات لتعديل الفيديو
-frame_count = st.slider("Number of frames", min_value=10, max_value=50, value=25, step=5)
-
 if uploaded_image is not None:
-    image = Image.open(uploaded_image)
+    image = Image.open(uploaded_image).convert("RGB")
     st.image(image, caption='Uploaded Image', use_column_width=True)
 
-    if st.button('Generate Video'):
-        # تحويل الصورة إلى فيديو
-        with st.spinner("Generating video..."):
-            video_frames = pipe(image, num_frames=frame_count)
+    if st.button('Generate 3D Mesh'):
+        # تحويل الصورة إلى تنسور
+        image_tensor = torch.tensor(np.array(image)).float().unsqueeze(0).permute(0, 3, 1, 2).to("cuda") / 255.0
         
-        st.success("Video generated successfully!")
+        # توليد الشبكة ثلاثية الأبعاد
+        with st.spinner("Generating 3D mesh..."):
+            mesh = pipe(image_tensor).images
+
+        st.success("3D mesh generated successfully!")
 
-        # عرض الفيديو
-        st.video(video_frames[0], format="video/mp4")
+        # عرض النتيجة
+        st.write("### 3D Mesh Preview")
+        # حاوية لعرض الـ 3D Mesh
+        # ملاحظة: قد تحتاج إلى تكوين طريقة لعرض أو تحميل النموذج ثلاثي الأبعاد بشكل صحيح.
+        st.image(mesh[0], caption="Generated 3D Mesh", use_column_width=True)
 
-        # تنزيل الفيديو
-        video_bytes = io.BytesIO()
-        video_frames[0].save(video_bytes, format="mp4")
-        st.download_button(label="Download Video", data=video_bytes.getvalue(), file_name="generated_video.mp4", mime="video/mp4")
+        # تنزيل النموذج
+        mesh_bytes = io.BytesIO()
+        mesh[0].save(mesh_bytes, format="png")  # استخدم الصيغة المناسبة هنا
+        st.download_button(label="Download 3D Mesh", data=mesh_bytes.getvalue(), file_name="generated_mesh.png", mime="image/png")
 
 # تقديم بعض المعلومات حول النموذج
 st.write("""
-### About the Model:
-SVD Image-to-Video is a latent diffusion model trained to generate short video clips from an image conditioning. 
-This model generates frames at a resolution of 576x1024 given a context frame of the same size, fine-tuned from the SVD Image-to-Video [14 frames] model.
-The widely used f8-decoder is also fine-tuned for temporal consistency, making the output videos more stable and coherent.
-""")
+### About InstantMesh:
+InstantMesh is a feed-forward framework for instant 3D mesh generation from a single image. It uses a combination of a multiview diffusion model and a sparse-view reconstruction model based on the LRM architecture to create diverse 3D assets quickly. The framework integrates a differentiable iso-surface extraction module to optimize the mesh representation, making it highly efficient and accurate.
+""")