updated app.py

.gitattributes

@@ -33,3 +33,4 @@ 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
+video_sample.mp4 filter=lfs diff=lfs merge=lfs -text

app.py

@@ -1,76 +1,100 @@
 import gradio as gr
-from PIL import Image, ImageOps
+from PIL import Image
 import numpy as np
-from tibblingai import wta
-import matplotlib.pyplot as plt  # needed for plt.imread
-def load_sample_images():
-    sample_paths = ["sample1.jpg", "sample2.jpg"]  # Must be in the same folder as your script
-    original_imgs = [Image.open(path) for path in sample_paths]
-    processed_imgs = [process_image(path) for path in sample_paths]
-    return gr.update(visible=True, value=original_imgs), gr.update(visible=True, value=processed_imgs)
-
-
-
-# Image processing function (you can replace this)
-def process_image(img: Image.Image) -> Image.Image:
-    # Read image using matplotlib
-    img_np = plt.imread(img)
+import matplotlib.pyplot as plt
+from tibbtech import neuroai
+import tempfile
+import os
+import imageio
+from matplotlib import cm
+
+# Image processing
+def process_image(img_path: str):
+    img_np = plt.imread(img_path)
     if img_np.ndim == 3 and img_np.shape[2] > 3:
         img_np = img_np[:, :, :3]
 
-    # Convert to grayscale for original image display
-    gray_img = ImageOps.grayscale(Image.fromarray((img_np * 255).astype(np.uint8)))
-
-    # Run your WTA processing (dummy if not available)
-    # Replace this line with actual WTA processing
-    img_tensor = wta.wta(img_np).numpy()  # Assuming returns shape (1, H, W)
-
-    # Convert processed image to inferno colormap
+    img_tensor = neuroai.wta(img_np).numpy()  # (1,H,W)
     inferno_colored = plt.cm.inferno(img_tensor[0])
     inferno_img = Image.fromarray((inferno_colored[:, :, :3] * 255).astype(np.uint8))
 
-    return gray_img, inferno_img
+    return img_path, inferno_img  # return path for original gallery
 
-# Function to process uploaded images
-def process_images(images):
-    if not isinstance(images, list):
-        images = [images]
+# Video processing with imageio (more browser-compatible)
+def process_video(video_path: str):
+    video_tensor = neuroai.run_wta(video_path)  # (T,H,W)
+    T, H, W = video_tensor.shape
 
-    original_imgs = []
-    processed_imgs = []
+    temp_path = tempfile.mktemp(suffix=".mp4")
 
-    for img in images:
-        original_gray_scale_img, inferno_img = process_image(img)
-        original_imgs.append(original_gray_scale_img)
-        processed_imgs.append(inferno_img)
+    # Normalize video to [0,1]
+    video_tensor = np.clip(video_tensor, 0, 1)
 
-    return gr.update(visible=True, value=original_imgs), gr.update(visible=True, value=processed_imgs)
+    writer = imageio.get_writer(
+        temp_path, fps=30, codec="libx264", ffmpeg_params=["-pix_fmt", "yuv420p"]
+    )
 
-# Gradio Interface
-with gr.Blocks() as demo:
-    gr.Markdown("## Upload Image(s) for Processing")
+    cmap = cm.get_cmap("inferno")
+    for t in range(T):
+        frame = video_tensor[t]
+        colored = (cmap(frame)[..., :3] * 255).astype(np.uint8)
+        writer.append_data(colored)
+    writer.close()
 
-    # File component (no preview shown here)
-    file_input = gr.File(file_types=["image"],file_count="multiple", label="Upload Images")
+    return video_path, temp_path
 
-    with gr.Row():
-        original_gallery = gr.Gallery(label="Original Images", visible=False)
-        processed_gallery = gr.Gallery(label="Processed Images", visible=False)
+# Unified processing
+def process_files(files):
+    if not isinstance(files, list):
+        files = [files]
 
-    file_input.change(fn=process_images, inputs=file_input, outputs=[original_gallery, processed_gallery])
-    with gr.Row():
-      with gr.Column(scale=0, min_width=300):
-        load_examples_btn = gr.Button("Run WTA on sample images")
-        load_examples_btn.click(fn=load_sample_images, outputs=[original_gallery, processed_gallery])
+    originals = []
+    processed = []
+
+    for file in files:
+        ext = os.path.splitext(file.name)[1].lower()
+        if ext in [".jpg", ".jpeg", ".png", ".bmp", ".tiff"]:
+            orig, proc = process_image(file.name)
+            originals.append(orig)
+            processed.append(proc)
+        elif ext in [".mp4", ".avi", ".mov", ".mkv"]:
+            orig, proc = process_video(file.name)
+            originals.append(orig)
+            processed.append(proc)
+
+    return (
+        gr.update(visible=bool(originals), value=originals),
+        gr.update(visible=bool(processed), value=processed)
+    )
 
+# Gradio interface
+with gr.Blocks() as demo:
+    gr.Markdown("## Upload Image(s) or Video(s) for Processing")
+
+    file_input = gr.File(file_types=["image", "video"], label="Upload Images/Videos", )
 
-    # 🆕 Examples section
-#     examples = gr.Examples(
-#     examples=[["sample1.jpg"], ["sample2.jpg"]],
-#     inputs=file_input,
-#     label="Example Images"
-# )
+    with gr.Row():
+        original_gallery = gr.Gallery(label="Original", visible=False)
+        processed_gallery = gr.Gallery(label="Processed", visible=False)
+
+    file_input.change(
+        fn=process_files,
+        inputs=file_input,
+        outputs=[original_gallery, processed_gallery]
+    )
+
+    example_files = [
+        ["sample1.jpg"],       # image
+        ["sample2.jpg"],       # image
+        ["video_sample.mp4"]   # video
+    ]
+
+    gr.Examples(
+        examples=example_files,
+        inputs=file_input,
+        fn=process_files,
+        outputs=[original_gallery, processed_gallery],
+        label="Try our sample images or video"
+    )
 
-    # example_files = [["sample1.jpg", "sample2.jpg"], ["sample2.jpg"]]
-    # gr.Examples(examples=example_files, inputs=[file_input], label="Try one of our example samples")
 demo.launch(debug=True)

video_sample.mp4

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2783b6f1f3a78a08086512d50babc8f5d90fa0e630ec7ef49acada9be759af4d
+size 242175