import sys # System package for path dependencies sys.path.append('Engineering-Clinic-Emerging-AI-Design-Interface/Interface_Dependencies') sys.path.append('Engineering-Clinic-Emerging-AI-Design-Interface/Interface') sys.path.append('Engineering-Clinic-Emerging-AI-Design-Interface/yolov7-main') sys.path.append('Engineering-Clinic-Emerging-AI-Design-Interface') sys.path.append('Interface_Dependencies') sys.path.append('Interface') from run_methods import run_all, correct_video import gradio as gr # Gradio package for interface # Gradio Interface Code def build_detect_interface(): with gr.Blocks(title="yolov7 Interface",theme=gr.themes.Base()) as demo: gr.Markdown( """ # Image & Video Interface for yolov7 Model ### Upload your own image or video and watch yolov7 try to guess what it is! """) # Row for for input & output settings with gr.Row() as file_settings: # Allows the user to toggle between the advanced and simple interfaces skill_type = gr.Radio(label="Skill Type",info="Choose 'Advanced' if you would like to use the advanced interface, Choose 'Simple' if you would like that version", choices=['Advanced','Simple', 'Help'],value='Simple',show_label=True,interactive=True,visible=True) # Allows choice for uploading image or video [for all] file_type = gr.Radio(label="File Type",info="Choose 'Image' if you are uploading an image, Choose 'Video' if you are uploading a video", choices=['Image','Video'],value='Image',show_label=True,interactive=True,visible=True) # Allows choice of source, from computer or webcam [for all] source_type = gr.Radio(label="Source Type",info="Choose 'Computer' if you are uploading from your computer, Choose 'Webcam' if you would like to use your webcam", choices=['Computer','Webcam'],value='Computer',show_label=True,interactive=True,visible=True) # Allows choice of which convolutional layer to show (1-17) [only for images] conv_layer = gr.Slider(label="Convolution Layer",info="Choose a whole number from 1 to 17 to see the corresponding convolutional layer", minimum=1,maximum=17,value=1,interactive=True,step=1,show_label=True) # Allows choice if video from webcam is streaming or uploaded [only for webcam videos] video_stream = gr.Checkbox(label="Stream from webcam?",info="Check this box if you would like to stream from your webcam",value=False,show_label=True,interactive=True,visible=False) # Allows choice of which smooth gradient output to show (1-3) [only for images] output_map = gr.Slider(label="Map Output Number",info="Choose a whole number from 1 to 3 to see the corresponding attribution map", minimum=1,maximum=3,value=1,interactive=True,step=1,show_label=True) # Row for all inputs & outputs with gr.Row() as inputs_outputs: # Default input image: Visible, Upload from computer input_im = gr.Image(source="upload",type='filepath',label="Input Image", show_download_button=True,show_share_button=True,interactive=True,visible=False) # Default Boxed output image: Visible output_box_im = gr.Image(type='filepath',label="Output Image", show_download_button=True,show_share_button=True,interactive=False,visible=False) # Defualt Convolutional output image: Visible output_conv_im = gr.Image(type='filepath',label="Output Convolution", show_download_button=True,show_share_button=True,interactive=False,visible=False) # Default Gradient output image: Visible output_grad_im = gr.Image(type='filepath',label="Output Smooth Gradient", show_download_button=True,show_share_button=True,interactive=False,visible=False) # Default label output textbox: Visible labels = gr.Textbox(label='Top Predictions:', value = "") # Default plaus output textbox: Visible plaus = gr.Textbox(label = "Plausibility Score:", value="") # Default time output textbox: Visible formatted_time = gr.Textbox(label = 'Total and Detection Time in Seconds:', value = "",scale=6) # Default input video: Not visible, Upload from computer input_vid = gr.Video(source="upload",label="Input Video", show_share_button=True,interactive=True,visible=False) # Default Boxed output video: Not visible output_box_vid = gr.Video(label="Output Video",show_share_button=True,visible=False) #An option to bring up a help menu help_box = gr.Textbox(lines=5, label="Help Box", visible=False) # List of components for clearing clear_comp_list = [input_im, output_box_im, output_conv_im, output_grad_im, labels, plaus, formatted_time, input_vid, output_box_vid] # Row for start, clear and demo buttons with gr.Row() as buttons: start_but = gr.Button(value="Start") demo1_but = gr.Button(value="ImageDemo") demo2_but = gr.Button(value="VideoDemo") clear_but = gr.ClearButton(value='Clear All',components=clear_comp_list, interactive=True,visible=True) # Row for model settings with gr.Row() as model_settings: # Pixel size of the inference [Possibly useless, may remove] inf_size = gr.Number(label='Inference Size (pixels)',value=640,precision=0,visible=False) # Object confidence threshold obj_conf_thr = gr.Number(label='Object Confidence Threshold',value=0.25,visible=False) # Intersection of union threshold iou_thr = gr.Number(label='IOU threshold for NMS',value=0.45,visible=False) # Agnostic NMS boolean agnostic_nms = gr.Checkbox(label='Agnostic NMS',value=True, visible=False) # Normailze gradient boolean norm = gr.Checkbox(label='Normalize Gradient',value=False,visible=False) # Weights File Upload weights = gr.File(label='Weights File',type='file',file_count='single',file_types=["pt"],value="weights/yolov7.pt",visible=False) #Consider making this its own file def change_file_type(file, source, is_stream, skill_type): """ Changes the visible components of the gradio interface Args: file (str): Type of the file (image or video) source (str): If the file is uploaded or from webcam is_stream (bool): If the video is streaming or uploaded Returns: Dictionary: Each component of the interface that needs to be updated. """ help_string = ("Variables within the interface: \n \ Simple Mode: \n \ Skill Type - The mode the interface will be set to. Simple is recommended for new users! \n \ File Type - Set the interface for image or video input \n \ Source Type - Set the input for media to computer(will allow uploads) or webcam(will use your devices webcam) \n \ stream from webcam? - Toggle the ability to livestream video from webcam \n \ Start - Run the interface with the set parameters and media \n \ ImageDemo - Run a demo to test the interface with a stock image \n \ VideoDemo - Run a demo to test the interface with a stock video \n \ ClearAll - Clear all of the input/output figures on the interface \n \ Advanced Mode: \n \ Convolution Layer - The convolutional layor extracted from the neural network. There are 17 of these to scroll between. \n \ Map Output Number - The attribution map. 1 is bounding box, 2 is object, and 3 is class. Currently, this is from the model not loss. \n \ Top Predictions - A list of the predictions along with their confidence percentage \n \ Plausibility Score - The measurement of how much the predicted bounding boxes overlap with the attributions generated (bounding box model output) \n \ Inference Size - The resolution of the image that will be interpreted (default is 640 by 640 pixels) \n \ Object Confidence Threshold - The minimum confidence required to display a bounding box \n \ IOU Threshold for NMS - The minimum confidence required for a bounding box to surpass another in cases of multiple predictions \n \ Agnostic NMS - Enable non max suppression to use the bounding boxes with the highest confidence in prediction \n \ Normalize Gradient - Scale the gradient vector to have a magnitude of 1 \n \ Weights File - A .pt file that contains the weight values that make the filter. This contains the classes and will greatly affect how an image or video is classified") if file == "Image": if source == "Computer": if skill_type == "Advanced": return { conv_layer: gr.Slider(visible=True), video_stream: gr.Checkbox(visible=False, value=False), output_map: gr.Slider(visible=True), input_im: gr.Image(source="upload",type='filepath',label="Input Image", show_download_button=True,show_share_button=True,interactive=True,visible=True,streaming=False,scale=3), output_box_im: gr.Image(visible=True,scale=3), output_conv_im: gr.Image(visible=True,scale=3), output_grad_im: gr.Image(visible=True,scale=3), input_vid: gr.Video(visible=False), output_box_vid: gr.Video(visible=False), norm: gr.Checkbox(visible=True), labels: gr.Textbox(visible=True,scale=1), plaus: gr.Textbox(visible=True,scale=1), formatted_time: gr.Textbox(visible=True,scale=1), help_box: gr.Textbox(visible=False, value=help_string), inf_size: gr.Number(visible=True), obj_conf_thr: gr.Number(visible=True), iou_thr: gr.Number(visible=True), agnostic_nms: gr.Checkbox(visible=True), weights: gr.File(visible=True) } elif skill_type == "Simple": return { conv_layer: gr.Slider(visible=False), video_stream: gr.Checkbox(visible=False, value=False), output_map: gr.Slider(visible=False), input_im: gr.Image(source="upload",type='filepath',label="Input Image", show_download_button=True,show_share_button=True,interactive=True,visible=True,streaming=False,scale=3), output_box_im: gr.Image(visible=True,scale=3), output_conv_im: gr.Image(visible=False,scale=3), output_grad_im: gr.Image(visible=False,scale=3), input_vid: gr.Video(visible=False), output_box_vid: gr.Video(visible=False), norm: gr.Checkbox(visible=False), labels: gr.Textbox(visible=False,scale=1), plaus: gr.Textbox(visible=False,scale=1), formatted_time: gr.Textbox(visible=False,scale=1), help_box: gr.Textbox(visible=False, value=help_string), inf_size: gr.Number(visible=False), obj_conf_thr: gr.Number(visible=False), iou_thr: gr.Number(visible=False), agnostic_nms: gr.Checkbox(visible=False), weights: gr.File(visible=False) } # Else its help else: return { conv_layer: gr.Slider(visible=False), video_stream: gr.Checkbox(visible=False, value=False), output_map: gr.Slider(visible=False), input_im: gr.Image(source="upload",type='filepath',label="Input Image", show_download_button=True,show_share_button=True,interactive=True,visible=False,streaming=False,scale=3), output_box_im: gr.Image(visible=False,scale=3), output_conv_im: gr.Image(visible=False,scale=3), output_grad_im: gr.Image(visible=False,scale=3), input_vid: gr.Video(visible=False), output_box_vid: gr.Video(visible=False), norm: gr.Checkbox(visible=False), labels: gr.Textbox(visible=False,scale=1), plaus: gr.Textbox(visible=False,scale=1), formatted_time: gr.Textbox(visible=False,scale=1), help_box: gr.Textbox(visible=True, value=help_string,scale=10), inf_size: gr.Number(visible=False), obj_conf_thr: gr.Number(visible=False), iou_thr: gr.Number(visible=False), agnostic_nms: gr.Checkbox(visible=False), weights: gr.File(visible=False) } elif source == "Webcam": if skill_type == "Advanced": return { conv_layer: gr.Slider(visible=True), video_stream: gr.Checkbox(visible=False, value=False), output_map: gr.Slider(visible=True), input_im: gr.Image(type='pil',source="webcam",label="Input Image", visible=True,interactive=True,streaming=False,scale=3), output_box_im: gr.Image(visible=True,scale=3), output_conv_im: gr.Image(visible=True,scale=3), output_grad_im: gr.Image(visible=True,scale=3), input_vid: gr.Video(visible=False), output_box_vid: gr.Video(visible=False), norm: gr.Checkbox(visible=True), labels: gr.Textbox(visible=True), plaus: gr.Textbox(visible=True), formatted_time: gr.Textbox(visible=True), help_box: gr.Textbox(visible=False, value=help_string), inf_size: gr.Number(visible=True), obj_conf_thr: gr.Number(visible=True), iou_thr: gr.Number(visible=True), agnostic_nms: gr.Checkbox(visible=True), weights: gr.File(visible=True) } elif skill_type == "Simple": return { conv_layer: gr.Slider(visible=False), video_stream: gr.Checkbox(visible=False, value=False), output_map: gr.Slider(visible=False), input_im: gr.Image(type='pil',source="webcam",label="Input Image", visible=True,interactive=True,streaming=False,scale=3), output_box_im: gr.Image(visible=True,scale=3), output_conv_im: gr.Image(visible=False,scale=3), output_grad_im: gr.Image(visible=False,scale=3), input_vid: gr.Video(visible=False), output_box_vid: gr.Video(visible=False), norm: gr.Checkbox(visible=False), labels: gr.Textbox(visible=False), plaus: gr.Textbox(visible=False), formatted_time: gr.Textbox(visible=False), help_box: gr.Textbox(visible=False, value=help_string), inf_size: gr.Number(visible=False), obj_conf_thr: gr.Number(visible=False), iou_thr: gr.Number(visible=False), agnostic_nms: gr.Checkbox(visible=False), weights: gr.File(visible=False) } else: return { conv_layer: gr.Slider(visible=False), video_stream: gr.Checkbox(visible=False, value=False), output_map: gr.Slider(visible=False), input_im: gr.Image(source="upload",type='filepath',label="Input Image", show_download_button=True,show_share_button=True,interactive=True,visible=False,streaming=False,scale=3), output_box_im: gr.Image(visible=False,scale=3), output_conv_im: gr.Image(visible=False,scale=3), output_grad_im: gr.Image(visible=False,scale=3), input_vid: gr.Video(visible=False), output_box_vid: gr.Video(visible=False), norm: gr.Checkbox(visible=False), labels: gr.Textbox(visible=False,scale=1), plaus: gr.Textbox(visible=False,scale=1), formatted_time: gr.Textbox(visible=False,scale=1), help_box: gr.Textbox(visible=True, value=help_string,scale=10), inf_size: gr.Number(visible=False), obj_conf_thr: gr.Number(visible=False), iou_thr: gr.Number(visible=False), agnostic_nms: gr.Checkbox(visible=False), weights: gr.File(visible=False) } elif file == "Video": if source == "Computer": if skill_type == "Advanced": return { conv_layer: gr.Slider(visible=False), video_stream: gr.Checkbox(visible=False, value=False), output_map: gr.Slider(visible=False), input_im: gr.Image(visible=False,streaming=False), output_box_im: gr.Image(visible=False), output_conv_im: gr.Image(visible=False), output_grad_im: gr.Image(visible=False), input_vid: gr.Video(source="upload",label="Input Video", show_share_button=True,interactive=True,visible=True,scale=6), output_box_vid: gr.Video(label="Output Video",show_share_button=True,visible=True,scale=6), norm: gr.Checkbox(visible=True), labels: gr.Textbox(visible=False), plaus: gr.Textbox(visible=False), formatted_time: gr.Textbox(visible=True,scale=1), help_box: gr.Textbox(visible=False, value=help_string,scale=10), inf_size: gr.Number(visible=True), obj_conf_thr: gr.Number(visible=True), iou_thr: gr.Number(visible=True), agnostic_nms: gr.Checkbox(visible=True), weights: gr.File(visible=True) } elif skill_type == "Simple": return { conv_layer: gr.Slider(visible=False), video_stream: gr.Checkbox(visible=False, value=False), output_map: gr.Slider(visible=False), input_im: gr.Image(visible=False,streaming=False), output_box_im: gr.Image(visible=False), output_conv_im: gr.Image(visible=False), output_grad_im: gr.Image(visible=False), input_vid: gr.Video(source="upload",label="Input Video", show_share_button=True,interactive=True,visible=True,scale=6), output_box_vid: gr.Video(label="Output Video",show_share_button=True,visible=True,scale=6), norm: gr.Checkbox(visible=False), labels: gr.Textbox(visible=False), plaus: gr.Textbox(visible=False), formatted_time: gr.Textbox(visible=False), help_box: gr.Textbox(visible=False, value=help_string), inf_size: gr.Number(visible=False), obj_conf_thr: gr.Number(visible=False), iou_thr: gr.Number(visible=False), agnostic_nms: gr.Checkbox(visible=False), weights: gr.File(visible=False) } else: return { conv_layer: gr.Slider(visible=False), video_stream: gr.Checkbox(visible=False, value=False), output_map: gr.Slider(visible=False), input_im: gr.Image(source="upload",type='filepath',label="Input Image", show_download_button=True,show_share_button=True,interactive=True,visible=False,streaming=False,scale=3), output_box_im: gr.Image(visible=False,scale=3), output_conv_im: gr.Image(visible=False,scale=3), output_grad_im: gr.Image(visible=False,scale=3), input_vid: gr.Video(visible=False), output_box_vid: gr.Video(visible=False), norm: gr.Checkbox(visible=False), labels: gr.Textbox(visible=False,scale=1), plaus: gr.Textbox(visible=False,scale=1), formatted_time: gr.Textbox(visible=False,scale=1), help_box: gr.Textbox(visible=True, value=help_string,scale=10), inf_size: gr.Number(visible=False), obj_conf_thr: gr.Number(visible=False), iou_thr: gr.Number(visible=False), agnostic_nms: gr.Checkbox(visible=False), weights: gr.File(visible=False) } #TODO-clinic: Fix possible streaming issues elif source == "Webcam": if is_stream: if skill_type == "Advanced": return { conv_layer: gr.Slider(visible=False), video_stream: gr.Checkbox(visible=True), output_map: gr.Slider(visible=False), input_im: gr.Image(type='pil',source="webcam",label="Input Image", streaming=True,visible=True,interactive=True,scale=6), output_box_im: gr.Image(visible=True,scale=6), output_conv_im: gr.Image(visible=False), #TODO-clinic: We may not want to hide all of these output_grad_im: gr.Image(visible=False), input_vid: gr.Video(visible=False), output_box_vid: gr.Video(visible=False), norm: gr.Checkbox(visible=True), labels: gr.Textbox(visible=False), plaus: gr.Textbox(visible=False), formatted_time: gr.Textbox(visible=True,scale=1), help_box: gr.Textbox(visible=False, value=help_string), inf_size: gr.Number(visible=True), obj_conf_thr: gr.Number(visible=True), iou_thr: gr.Number(visible=True), agnostic_nms: gr.Checkbox(visible=True), weights: gr.File(visible=True) } elif skill_type == "Simple": return { conv_layer: gr.Slider(visible=False), video_stream: gr.Checkbox(visible=True), output_map: gr.Slider(visible=False), input_im: gr.Image(type='pil',source="webcam",label="Input Image", streaming=True,visible=True,interactive=True,scale=6), output_box_im: gr.Image(visible=True,scale=6), output_conv_im: gr.Image(visible=False), output_grad_im: gr.Image(visible=False), input_vid: gr.Video(visible=False), output_box_vid: gr.Video(visible=False), norm: gr.Checkbox(visible=False), labels: gr.Textbox(visible=False), plaus: gr.Textbox(visible=False), formatted_time: gr.Textbox(visible=False), help_box: gr.Textbox(visible=False, value=help_string), inf_size: gr.Number(visible=False), obj_conf_thr: gr.Number(visible=False), iou_thr: gr.Number(visible=False), agnostic_nms: gr.Checkbox(visible=False), weights: gr.File(visible=False) } else: return { conv_layer: gr.Slider(visible=False), video_stream: gr.Checkbox(visible=False, value=False), output_map: gr.Slider(visible=False), input_im: gr.Image(source="upload",type='filepath',label="Input Image", show_download_button=True,show_share_button=True,interactive=True,visible=False,streaming=False,scale=3), output_box_im: gr.Image(visible=False,scale=3), output_conv_im: gr.Image(visible=False,scale=3), output_grad_im: gr.Image(visible=False,scale=3), input_vid: gr.Video(visible=False), output_box_vid: gr.Video(visible=False), norm: gr.Checkbox(visible=False), labels: gr.Textbox(visible=False,scale=1), plaus: gr.Textbox(visible=False,scale=1), formatted_time: gr.Textbox(visible=False,scale=1), help_box: gr.Textbox(visible=True, value=help_string,scale=10), inf_size: gr.Number(visible=False), obj_conf_thr: gr.Number(visible=False), iou_thr: gr.Number(visible=False), agnostic_nms: gr.Checkbox(visible=False), weights: gr.File(visible=False) } elif not is_stream: if skill_type == "Advanced": return { #TODO-clinic: We may not want to hide all of these conv_layer: gr.Slider(visible=False), video_stream: gr.Checkbox(visible=True, value=False), output_map: gr.Slider(visible=False), input_im: gr.Image(visible=False,streaming=False), output_box_im: gr.Image(visible=False), output_conv_im: gr.Image(visible=False), output_grad_im: gr.Image(visible=False), input_vid: gr.Video(label="Input Video",source="webcam", show_share_button=True,interactive=True,visible=True,scale=6), output_box_vid: gr.Video(label="Output Video",show_share_button=True,visible=True,scale=6), norm: gr.Checkbox(visible=True), labels: gr.Textbox(visible=False), plaus: gr.Textbox(visible=False), formatted_time: gr.Textbox(visible=True,scale=1), help_box: gr.Textbox(visible=False, value=help_string), inf_size: gr.Number(visible=True), obj_conf_thr: gr.Number(visible=True), iou_thr: gr.Number(visible=True), agnostic_nms: gr.Checkbox(visible=True), weights: gr.File(visible=True) } elif skill_type == "Simple": #TODO-clinic: Update for simple return { conv_layer: gr.Slider(visible=False), video_stream: gr.Checkbox(visible=True, value=False), output_map: gr.Slider(visible=False), input_im: gr.Image(visible=False,streaming=False), output_box_im: gr.Image(visible=False), output_conv_im: gr.Image(visible=False), output_grad_im: gr.Image(visible=False), input_vid: gr.Video(label="Input Video",source="webcam", show_share_button=True,interactive=True,visible=True,scale=6), output_box_vid: gr.Video(label="Output Video",show_share_button=True,visible=True,scale=6), norm: gr.Checkbox(visible=False), labels: gr.Textbox(visible=False), plaus: gr.Textbox(visible=False), formatted_time: gr.Textbox(visible=False), help_box: gr.Textbox(visible=False, value=help_string), inf_size: gr.Number(visible=False), obj_conf_thr: gr.Number(visible=False), iou_thr: gr.Number(visible=False), agnostic_nms: gr.Checkbox(visible=False), weights: gr.File(visible=False) } else: return { conv_layer: gr.Slider(visible=False), video_stream: gr.Checkbox(visible=False, value=False), output_map: gr.Slider(visible=False), input_im: gr.Image(source="upload",type='filepath',label="Input Image", show_download_button=True,show_share_button=True,interactive=True,visible=False,streaming=False,scale=3), output_box_im: gr.Image(visible=False,scale=3), output_conv_im: gr.Image(visible=False,scale=3), output_grad_im: gr.Image(visible=False,scale=3), input_vid: gr.Video(visible=False), output_box_vid: gr.Video(visible=False), norm: gr.Checkbox(visible=False), labels: gr.Textbox(visible=False,scale=1), plaus: gr.Textbox(visible=False,scale=1), formatted_time: gr.Textbox(visible=False,scale=1), help_box: gr.Textbox(visible=True, value=help_string,scale=10), inf_size: gr.Number(visible=False), obj_conf_thr: gr.Number(visible=False), iou_thr: gr.Number(visible=False), agnostic_nms: gr.Checkbox(visible=False), weights: gr.File(visible=False) } def change_conv_layer(layer): """ Changes the shown convolutional output layer based on gradio slider Args: layer (int): The layer to show Returns: str: The file path of the output image """ return "outputs/runs/detect/exp/layers/layer" + str(int(int(layer) - 1)) + '.jpg' def change_output_num(number): """ Changes the shown gradient map based on gradio slider Args: number (int): The gradient map to show Returns: str: The file path of the output image """ return "outputs/runs/detect/exp/smoothGrad" + str(int(int(number) -1)) + '.jpg' def demo1(): """ Run a test image file as a demo """ #TODO-clinic - There may be a better way to do this (move to another file) file_type = gr.Radio(label="File Type",info="Choose 'Image' if you are uploading an image, Choose 'Video' if you are uploading a video", choices=['Image','Video'],value='Image',show_label=True,interactive=True,visible=False) input_im = gr.Image(type='filepath',label="Input Image",value= "references/inference/images/bus.jpg", show_download_button=True,show_share_button=True,interactive=True,visible=False) input_vid = gr.Video(source="upload",label="Input Video", show_share_button=True,interactive=True,visible=False) source_type = gr.Radio(label="Source Type",info="Choose 'Computer' if you are uploading from your computer, Choose 'Webcam' if you would like to use your webcam", choices=['Computer','Webcam'],value='Computer',show_label=True,interactive=True,visible=False) inf_size = gr.Number(label='Inference Size (pixels)',value=640,precision=0,visible=False) obj_conf_thr = gr.Number(label='Object Confidence Threshold',value=0.25,visible=False) iou_thr = gr.Number(label='IOU threshold for NMS',value=0.45,visible=False) conv_layer = gr.Slider(label="Convolution Layer",info="Choose a whole number from 1 to 17 to see the corresponding convolutional layer", minimum=1,maximum=17,value=1,interactive=True,step=1,show_label=True,visible=False) agnostic_nms = gr.Checkbox(label='Agnostic NMS',value=True,visible=False) output_map = gr.Slider(label="Map Output Number",info="Choose a whole number from 1 to 3 to see the corresponding attribution map", minimum=1,maximum=3,value=1,interactive=True,step=1,show_label=True,visible=False) video_stream = gr.Checkbox(label="Stream from webcam?",info="Check this box if you would like to stream from your webcam",value=False,show_label=True,interactive=True,visible=False) norm = gr.Checkbox(label='Normalize Gradient',value=False,visible=False) weights = gr.File(label='Weights File',type='file',file_count='single',file_types=["pt"],value="weights/yolov7.pt",visible=False) skill_type = gr.Radio(label="Skill Type",info="Choose 'Advanced' if you would like to use the advanced interface, 'Simple' if you would like that version, or help for more info", choices=['Advanced','Simple'],value='Advanced',show_label=True,interactive=True,visible=False) demo1_inputs = [file_type, input_im, input_vid, source_type, inf_size, obj_conf_thr, iou_thr, conv_layer, agnostic_nms, output_map, video_stream, norm, weights,skill_type] return demo1_inputs def demo2(): """ Run a test video file as a demo """ #TODO-clinic - There may be a better way to do this (move to another file) file_type = gr.Radio(label="File Type",info="Choose 'Image' if you are uploading an image, Choose 'Video' if you are uploading a video", choices=['Image','Video'],value='Video',show_label=True,interactive=True,visible=False) input_im = gr.Image(type='filepath',label="Input Image", show_download_button=True,show_share_button=True,interactive=True,visible=False) input_vid = gr.Video(source="upload",label="Input Video", show_share_button=True,interactive=True,visible=False,value= "references/inference/videos/ducks.mp4") source_type = gr.Radio(label="Source Type",info="Choose 'Computer' if you are uploading from your computer, Choose 'Webcam' if you would like to use your webcam", choices=['Computer','Webcam'],value='Computer',show_label=True,interactive=True,visible=False) inf_size = gr.Number(label='Inference Size (pixels)',value=640,precision=0,visible=False) obj_conf_thr = gr.Number(label='Object Confidence Threshold',value=0.25,visible=False) iou_thr = gr.Number(label='IOU threshold for NMS',value=0.45,visible=False) conv_layer = gr.Slider(label="Convolution Layer",info="Choose a whole number from 1 to 17 to see the corresponding convolutional layer", minimum=1,maximum=17,value=1,interactive=True,step=1,show_label=True,visible=False) agnostic_nms = gr.Checkbox(label='Agnostic NMS',value=True,visible=False) output_map = gr.Slider(label="Map Output Number",info="Choose a whole number from 1 to 3 to see the corresponding attribution map", minimum=1,maximum=3,value=1,interactive=True,step=1,show_label=True,visible=False) video_stream = gr.Checkbox(label="Stream from webcam?",info="Check this box if you would like to stream from your webcam",value=False,show_label=True,interactive=True,visible=False) norm = gr.Checkbox(label='Normalize Gradient',value=False,visible=False) weights = gr.File(label='Weights File',type='file',file_count='single',file_types=["pt"],value="weights/yolov7.pt",visible=False) skill_type = gr.Radio(label="Skill Type",info="Choose 'Advanced' if you would like to use the advanced interface, Choose 'Simple' if you would like that version", choices=['Advanced','Simple'],value='Advanced',show_label=True,interactive=True,visible=False) demo2_inputs = [file_type, input_im, input_vid, source_type, inf_size, obj_conf_thr, iou_thr, conv_layer, agnostic_nms, output_map, video_stream, norm, weights,skill_type] return demo2_inputs # List of gradio components that change during method "change_file_type" change_comp_list = [conv_layer, video_stream, output_map, input_im, output_box_im, output_conv_im, output_grad_im, input_vid, output_box_vid, norm, labels, plaus, formatted_time, help_box, inf_size, obj_conf_thr, iou_thr, agnostic_nms, weights] # List of gradio components that are input into the run_all method (when start button is clicked) run_inputs = [file_type, input_im, input_vid, source_type, inf_size, obj_conf_thr, iou_thr, conv_layer, agnostic_nms, output_map, video_stream, norm, weights, skill_type] #TODO-clinic: Skill_type might be able to be removed here # List of gradio components that are output from the run_all method (when start button is clicked) run_outputs = [output_box_im, output_conv_im, output_grad_im, labels, plaus, formatted_time, output_box_vid, input_im, input_vid] # When these settings are changed, the change_file_type method is called file_type.input(change_file_type, show_progress=True, inputs=[file_type, source_type, video_stream, skill_type], outputs=change_comp_list) source_type.input(change_file_type, show_progress=True, inputs=[file_type, source_type, video_stream, skill_type], outputs=change_comp_list) video_stream.input(change_file_type, show_progress=True, inputs=[file_type, source_type, video_stream, skill_type], outputs=change_comp_list) skill_type.input(change_file_type, show_progress=True, inputs=[file_type, source_type, video_stream, skill_type], outputs=change_comp_list) # When start button is clicked, the run_all method is called start_but.click(run_all, inputs=run_inputs, outputs=run_outputs) # When demo1 button is clicked, run the function with stored params demo1_but.click(run_all, inputs=demo1(), outputs=run_outputs) # # When demo2 button is clicked, run the function with stored params demo2_but.click(run_all, inputs=demo2(), outputs=run_outputs) # When video is uploaded, the correct_video method is called input_vid.upload(correct_video, inputs=[input_vid], outputs=[input_vid]) # When the convolutional layer setting is changed, the change_conv_layer method is called conv_layer.input(change_conv_layer, conv_layer, output_conv_im) # When the stream setting is true, run the stream input_im.stream(run_all, inputs=run_inputs, outputs=run_outputs) # When the gradient number is changed, the change_output_num method is called output_map.input(change_output_num, output_map, output_grad_im) # When the demo is first started, run the change_file_type method to ensure default settings demo.load(change_file_type, show_progress=True, inputs=[file_type, source_type, video_stream, skill_type], outputs=change_comp_list) return demo # #Used for debugging # if __name__== "__main__" : # # If True, it launches Gradio interface # # If False, it runs without the interface (for better debugging) # if True: # # demo.queue().launch(share=True) # demo.queue().launch() # else: # #Test with yolo (weights and normal photo) # run_all('Image','references\\inference\\images\\bus.jpg', 'None', 'Computer', 640,0.25,0.45,1,True,1,False,False,"yolov7.pt", "Advanced") # #Test drone (weights and drone photo) TODO: Weight set may not be included in the repo # #run_all('Image','references\\inference\\images\\V_DRONE_113_34.jpg', 'None', 'Computer', 640,0.25,0.45,1,True,1,False,False,"drone_63.pt", "Advanced")