Autocrop off by default, added progress bar for batch prediction, made colors nice, fixed download button

app.py

@@ -122,7 +122,7 @@ current_model_type = None
 
 results_cache: dict[str, str] = {}
 current_image = None
-autocrop = True
+autocrop = False
 
 temp_files: list[str] = []
 all_images: list[str] = []
@@ -157,7 +157,7 @@ def display_model():
     """
     global current_model_type
     model_name = model_names[current_model_type]
-    return f"Current Model Type: {model_name}. Reupload model to change it."
+    return f"Current Model Type: {model_name}. Use dropdown on the right to change it."
 
 
 def clear():
@@ -229,9 +229,9 @@ def predict(img: str) -> gr.BarPlot:
     )
 
 
-def predict_all():
+def predict_all(progress_bar=gr.Progress()):
     global all_images, results_cache
-    for img in all_images:
+    for img in progress_bar.tqdm(all_images, desc="Running images"):
         current_image_name = Path(img).name
         img = Image.open(img).convert("RGB")
         if autocrop:
@@ -243,6 +243,7 @@ def predict_all():
             "Distribution": result,
             "Classification": {"Probability": prob, "Label": label},
         }
+    return "All images predicted successfully."
 
 
 def get_results_cache():
@@ -302,8 +303,9 @@ with gr.Blocks() as demo:
         gr.Textbox(
             "Only use this application on the following classes of nematodes: "
             + "Helicotylenchus, Hoplolaimus, Meloidogyne, Mesocriconema, "
-            "Pratylenchus, Trichodorus, and Tylenchorhynchus\n\n"
-            + "Only use images containing a single nematode.",
+            "Pratylenchus, Trichodorus, and Tylenchorhynchus.\n\n"
+            + "Only use images containing a single nematode.\n\n"
+            + "SCROLL DOWN TO DOWNLOAD THE PREDICTIONS FOR YOUR IMAGES!",
             text_align="center",
             label="DISCLAIMER",
         )
@@ -315,7 +317,7 @@ with gr.Blocks() as demo:
         model_select = gr.Dropdown(
             choices=["EfficientNetV2-S", "MobileNetV3-L", "ResNet101", "Swin V2-B"],
             value="EfficientNetV2-S",
-            label="Select Model Architecture",
+            label="Select Model Architecture (May take a few moments, check text on the left to confirm your model has loaded)",
         )
     with gr.Row():
         with gr.Column():
@@ -324,14 +326,17 @@ with gr.Blocks() as demo:
                 show_label=False,
             )
             files = gr.File(file_types=["image"], file_count="multiple")
-            batch_predict = gr.Button("Predict All")
+            batch_predict = gr.Button("Predict All", variant="stop")
+            prediction_progress = gr.Textbox(
+                "Prediction Progress Bar", show_label=False
+            )
 
         with gr.Column():
             mid_col_text = gr.Textbox(
                 "Crop Image Here (Optional), then Click Run to Predict",
                 show_label=False,
             )
-            autocrop_toggle = gr.Checkbox(value=True, label="Automatic Cropping")
+            autocrop_toggle = gr.Checkbox(value=False, label="Automatic Cropping")
             cropper = gr.ImageEditor(
                 type="filepath",
                 sources=None,
@@ -352,7 +357,7 @@ with gr.Blocks() as demo:
                 interactive=False,
                 mirror_webcam=False,
             )
-            classify = gr.Button("Classify")
+            classify = gr.Button("Classify", variant="stop")
             plot = gr.BarPlot()
 
     with gr.Row():
@@ -361,16 +366,15 @@ with gr.Blocks() as demo:
             label="Predictions",
         )
     with gr.Row():
-        json_results = gr.JSON()
-        download = gr.DownloadButton(
-            "Download Predictions (Click again if download does not start, Gradio bug)"
-        )
+        with gr.Column():
+            json_results = gr.JSON()
+        with gr.Column():
+            download = gr.DownloadButton("Download Predictions", variant="primary")
 
     download.click(save_results, outputs=download)
     model_select.change(load_model, inputs=model_select).then(
         display_model, outputs=model_text
     )
-    model_select
 
     files.upload(upload_files, inputs=files)
     files.select(select_image, inputs=files, outputs=cropper).then(
@@ -388,15 +392,19 @@ with gr.Blocks() as demo:
         outputs=preview,
     )
 
-    batch_predict.click(predict_all).then(get_results_cache, outputs=json_results)
+    batch_predict.click(predict_all, outputs=prediction_progress).then(
+        get_results_cache, outputs=json_results
+    ).then(save_results, outputs=download)
 
-    files.clear(clear).then(get_results_cache, outputs=json_results)
+    files.clear(clear).then(get_results_cache, outputs=json_results).then(
+        save_results, outputs=download
+    )
 
     crop.click(show_preview, inputs=cropper, outputs=preview)
 
     classify.click(predict, inputs=preview, outputs=plot).then(
         get_results_cache, outputs=json_results
-    )
+    ).then(save_results, outputs=download)
     demo.unload(clear)