Updated app to support latest gradio version

README.md

@@ -4,7 +4,7 @@ emoji: 🐢
 colorFrom: yellow
 colorTo: green
 sdk: gradio
-sdk_version: 3.28.1
+sdk_version: 4.7.1
 app_file: app.py
 pinned: false
 license: other

app.py

@@ -313,7 +313,7 @@ def predict(x,top_n_classes):
     labels = list(zip(sal_class_labels, class_conf[sal_class_idxs].tolist()))
     final_labels = dict(sorted(labels, key=lambda t: t[1],reverse=True)[:top_n_classes])
     
-    return final_labels, Dropdown.update(choices=list(final_labels))
+    return final_labels, Dropdown(choices=list(final_labels),label="Class to compute saliency",interactive=True,visible=True)
 
 # Interpretation function for image classification that implements the selected saliency algorithm and generates the class-wise saliency map visualizations
 def interpretation_function(image: np.ndarray, 
@@ -326,22 +326,18 @@ def interpretation_function(image: np.ndarray,
     
     sal_generator = get_saliency_algo(img_cls_saliency_algo_name[-1])
     sal_generator.fill = blackbox_fill
-    labels_list = [i['label'] for i in labels['confidences']]
+    labels_list = labels.keys()
     blackbox_classifier.set_labels(labels_list)
     sal_maps = sal_generator(image, blackbox_classifier)
     nth_class_index = blackbox_classifier.get_labels().index(nth_class)
-    scores = sal_maps[nth_class_index,:,:]
     fig = visualize_saliency_plot(image,
                                 sal_maps[nth_class_index,:,:],
                                 img_alpha,
                                 sal_alpha,
                                 sal_range_min,
                                 sal_range_max)
-    
-    scores = np.clip(scores, sal_range_min, sal_range_max)
         
-    return {"original": gr_processing_utils.encode_array_to_base64(image), 
-            "interpretation": scores.tolist()}, fig
+    return fig
 
 def visualize_saliency_plot(image: np.ndarray, 
                             class_sal_map: np.ndarray,
@@ -394,7 +390,7 @@ def run_detect(input_img: np.ndarray, num_detections: int):
     
     bboxes_list = bboxes[:,:].astype(int).tolist()
 
-    return (input_img, list(zip([f for f in bboxes_list], [l for l in final_label]))[:num_detections]), Dropdown.update(choices=[l for l in final_label][:num_detections])
+    return (input_img, list(zip([f for f in bboxes_list], [l for l in final_label]))[:num_detections]), Dropdown(choices=[l for l in final_label][:num_detections],label="Detection to compute saliency",interactive=True,visible=True)
 
 # Run saliency algorithm on the object detect predictions and generate corresponding visualizations
 def run_detect_saliency(input_img: np.ndarray, 
@@ -490,7 +486,7 @@ with gr.Blocks() as demo:
                 debiased = Checkbox(value=debiased_state[-1],label="Debiased", interactive=True, visible=False)
         with Row():
             with Column():
-                input_img = Image(label="Saliency Map Generation", shape=(640, 480))
+                input_img = Image(label="Saliency Map Generation", width=640, height=480)
                 num_classes = Slider(value=2,label="Top-N class labels", interactive=True,visible=True)
                 classify = Button("Classify")
             with Column():
@@ -510,8 +506,6 @@ with gr.Blocks() as demo:
                 with Tabs():
                     with TabItem("Display interpretation with plot"):
                         interpretation_plot = Plot()
-                    with TabItem("Display interpretation with built-in component"):
-                        interpretation = gr_components.Interpretation(input_img)
 
     with Tab("Object Detection"):
         with Row():
@@ -531,7 +525,7 @@ with gr.Blocks() as demo:
                 threads_detect = Slider(value=threads_state[-1],label="Threads",interactive=True,visible=False)
         with Row():
             with Column():
-                input_img_detect = Image(label="Saliency Map Generation", shape=(640, 480))
+                input_img_detect = Image(label="Saliency Map Generation", width=640, height=480)
                 num_detections = Slider(value=2,label="Top-N detections", interactive=True,visible=True)
                 detection = Button("Run Detection Algorithm")
             with Column():
@@ -572,7 +566,7 @@ with gr.Blocks() as demo:
     # Image Classification prediction and saliency generation event listeners
     classify.click(predict, [input_img, num_classes], [class_label,class_name])
     class_label.select(map_labels,None,class_name)
-    generate_saliency.click(interpretation_function, [input_img, class_label, class_name, img_alpha, sal_alpha, min_sal_range, max_sal_range], [interpretation,interpretation_plot])
+    generate_saliency.click(interpretation_function, [input_img, class_label, class_name, img_alpha, sal_alpha, min_sal_range, max_sal_range], [interpretation_plot])
 
     # Object Detection dropdown list event listeners
     drop_list_detect_model.select(select_obj_det_model,drop_list_detect_model,drop_list_detect_model)