update app.py

app.py

@@ -4,7 +4,7 @@ warnings.filterwarnings('ignore')
 
 import subprocess, io, os, sys, time
 
-os.system("pip install gradio==3.50.2")
+# os.system("pip install gradio==3.50.2")
 
 import gradio as gr
 from loguru import logger
@@ -123,6 +123,8 @@ ram_model = None
 kosmos_model = None
 kosmos_processor = None
 
+brush_color = "#00FFFF"
+
 def load_model_hf(model_config_path, repo_id, filename, device='cpu'):
     args = SLConfig.fromfile(model_config_path) 
     model = build_model(args)
@@ -593,6 +595,17 @@ def run_anything_task(input_image, text_prompt, task_type, inpaint_prompt, box_t
     run_task_time = 0
     time_cost_str = ''
     run_task_time, time_cost_str = get_time_cost(run_task_time, time_cost_str)
+        
+    logger.info(f"input_image==={input_image}")
+    if 'background' in input_image.keys():
+        input_image['image'] = input_image['background']
+        if len(input_image['layers']) > 0:
+            # input_image['mask'] = input_image['layers'][0]   #brush_color
+            img_arr = np.array(input_image['layers'][0].convert("L"))
+            logger.info(f"img_arr==={img_arr.shape}, {img_arr[760][640]}, {img_arr[0][0]}")
+            img_arr = np.where(img_arr > 0, 1, img_arr)
+            # img_arr = 1 - img_arr
+            input_image['mask'] = Image.fromarray(255*img_arr.astype('uint8'))
 
     if (task_type == 'Kosmos-2'):
         global kosmos_model, kosmos_processor
@@ -605,20 +618,20 @@ def run_anything_task(input_image, text_prompt, task_type, inpaint_prompt, box_t
             
         kosmos_image, kosmos_text, kosmos_entities = kosmos_generate_predictions(image_pil, kosmos_input, kosmos_model, kosmos_processor)
         run_task_time, time_cost_str = get_time_cost(run_task_time, time_cost_str)
-        return None, None, time_cost_str, kosmos_image, gr.Textbox.update(visible=(time_cost_str !='')), kosmos_text, kosmos_entities
+        return None, None, time_cost_str, kosmos_image, gr.update(visible=(time_cost_str !='')), kosmos_text, kosmos_entities
 
     if (task_type == 'relate anything'):
         output_images = relate_anything(input_image['image'], num_relation)
         run_task_time, time_cost_str = get_time_cost(run_task_time, time_cost_str)
-        return output_images, gr.Gallery.update(label='relate images'), time_cost_str, gr.Textbox.update(visible=(time_cost_str !='')), None, None, None
+        return output_images, gr.update(label='relate images'), time_cost_str, gr.update(visible=(time_cost_str !='')), None, None, None
 
     text_prompt = text_prompt.strip()
     if not ((task_type in ['inpainting', 'outpainting'] or task_type == 'remove') and mask_source_radio == mask_source_draw):
         if text_prompt == '':
-            return [], gr.Gallery.update(label='Detection prompt is not found!😂😂😂😂'), time_cost_str, gr.Textbox.update(visible=(time_cost_str !='')), None, None, None
+            return [], gr.update(label='Detection prompt is not found!😂😂😂😂'), time_cost_str, gr.update(visible=(time_cost_str !='')), None, None, None
 
     if input_image is None:
-            return [], gr.Gallery.update(label='Please upload a image!😂😂😂😂'), time_cost_str, gr.Textbox.update(visible=(time_cost_str !='')), None, None, None
+            return [], gr.update(label='Please upload a image!😂😂😂😂'), time_cost_str, gr.update(visible=(time_cost_str !='')), None, None, None
 
     file_temp = int(time.time())
     logger.info(f'run_anything_task_002/{device}_[{file_temp}]_{task_type}/{inpaint_mode}/[{mask_source_radio}]/{remove_mode}/{remove_mask_extend}_[{text_prompt}]/[{inpaint_prompt}]___1_')
@@ -661,7 +674,7 @@ def run_anything_task(input_image, text_prompt, task_type, inpaint_prompt, box_t
         )
         if boxes_filt.size(0) == 0:
             logger.info(f'run_anything_task_[{file_temp}]_{task_type}_[{text_prompt}]_1___{groundingdino_device}/[No objects detected, please try others.]_')
-            return [], gr.Gallery.update(label='No objects detected, please try others.😂😂😂😂'), time_cost_str, gr.Textbox.update(visible=(time_cost_str !='')), None, None, None
+            return [], gr.update(label='No objects detected, please try others.😂😂😂😂'), time_cost_str, gr.update(visible=(time_cost_str !='')), None, None, None
         boxes_filt_ori = copy.deepcopy(boxes_filt)
 
         pred_dict = {
@@ -726,7 +739,7 @@ def run_anything_task(input_image, text_prompt, task_type, inpaint_prompt, box_t
     logger.info(f'run_anything_task_[{file_temp}]_{task_type}_3_')
     if task_type == 'detection' or task_type == 'segment':
         logger.info(f'run_anything_task_[{file_temp}]_{task_type}_9_')
-        return output_images, gr.Gallery.update(label='result images'), time_cost_str, gr.Textbox.update(visible=(time_cost_str !='')), None, None, None
+        return output_images, gr.update(label='result images'), time_cost_str, gr.update(visible=(time_cost_str !='')), None, None, None
     elif task_type in ['inpainting', 'outpainting'] or task_type == 'remove':
         if inpaint_prompt.strip() == '' and mask_source_radio == mask_source_segment:
             task_type = 'remove'
@@ -804,11 +817,11 @@ def run_anything_task(input_image, text_prompt, task_type, inpaint_prompt, box_t
         output_images.append(image_inpainting)
         run_task_time, time_cost_str = get_time_cost(run_task_time, time_cost_str)
         logger.info(f'run_anything_task_[{file_temp}]_{task_type}_9_')
-        return output_images, gr.Gallery.update(label='result images'), time_cost_str, gr.Textbox.update(visible=(time_cost_str !='')), None, None, None       
+        return output_images, gr.update(label='result images'), time_cost_str, gr.update(visible=(time_cost_str !='')), None, None, None       
     else:
         logger.info(f"task_type:{task_type} error!")
     logger.info(f'run_anything_task_[{file_temp}]_9_9_')
-    return output_images, gr.Gallery.update(label='result images'), time_cost_str, gr.Textbox.update(visible=(time_cost_str !='')), None, None, None
+    return output_images, gr.update(label='result images'), time_cost_str, gr.update(visible=(time_cost_str !='')), None, None, None
 
 def change_radio_display(task_type, mask_source_radio):
     text_prompt_visible = True
@@ -839,14 +852,14 @@ def change_radio_display(task_type, mask_source_radio):
         text_prompt_visible = False
         num_relation_visible = True
 
-    return  (gr.Textbox.update(visible=text_prompt_visible), 
-            gr.Textbox.update(visible=inpaint_prompt_visible), 
-            gr.Radio.update(visible=mask_source_radio_visible), 
-            gr.Slider.update(visible=num_relation_visible),
-            gr.Gallery.update(visible=image_gallery_visible),
-            gr.Radio.update(visible=kosmos_input_visible),
-            gr.Image.update(visible=kosmos_output_visible),
-            gr.HighlightedText.update(visible=kosmos_text_output_visible))
+    return  (gr.update(visible=text_prompt_visible), 
+            gr.update(visible=inpaint_prompt_visible), 
+            gr.update(visible=mask_source_radio_visible), 
+            gr.update(visible=num_relation_visible),
+            gr.update(visible=image_gallery_visible),
+            gr.update(visible=kosmos_input_visible),
+            gr.update(visible=kosmos_output_visible),
+            gr.update(visible=kosmos_text_output_visible))
 
 def get_model_device(module):
     try:
@@ -883,9 +896,12 @@ def main_gradio(args):
                     task_types.append("relate anything")
                 if kosmos_enable:
                     task_types.append("Kosmos-2")           
-         
-                input_image = gr.Image(source='upload', elem_id="image_upload", tool='sketch', type='pil', label="Upload",
-                                    height=512, brush_color='#00FFFF', mask_opacity=0.6)    
+
+                # input_image = gr.Image(source='upload', elem_id="image_upload", tool='sketch', type='pil', label="Upload",
+                #                    height=512, brush_color='#00FFFF', mask_opacity=0.6) 
+                                    
+                input_image = gr.ImageMask(sources='upload', elem_id="image_upload", type='pil', label="Upload",
+                                    brush=gr.Brush(colors=[brush_color], color_mode="fixed"))
                 task_type = gr.Radio(task_types,  value="detection", 
                                                 label='Task type', visible=True) 
                 mask_source_radio = gr.Radio([mask_source_draw, mask_source_segment], 
@@ -897,7 +913,7 @@ def main_gradio(args):
                 
                 kosmos_input = gr.Radio(["Brief", "Detailed"], label="Kosmos Description Type", value="Brief", visible=False)
 
-                run_button = gr.Button(label="Run", visible=True)
+                run_button = gr.Button(value="Run", visible=True)
                 with gr.Accordion("Advanced options", open=False) as advanced_options:
                     box_threshold = gr.Slider(
                         label="Box Threshold", minimum=0.0, maximum=1.0, value=0.3, step=0.001
@@ -917,7 +933,7 @@ def main_gradio(args):
 
             with gr.Column():
                 image_gallery = gr.Gallery(label="result images", show_label=True, elem_id="gallery", height=512, visible=True
-                    ).style(preview=True, columns=[5], object_fit="scale-down", height="auto")   
+                    ) #.style(preview=True, columns=[5], object_fit="scale-down", height="auto")   
                 time_cost = gr.Textbox(label="Time cost by step (ms):", visible=False, interactive=False)
 
                 kosmos_output = gr.Image(type="pil", label="result images", visible=False)
@@ -926,9 +942,9 @@ def main_gradio(args):
                                     combine_adjacent=False,
                                     show_legend=True,
                                     visible=False,
-                                ).style(color_map=color_map)
+                                ) # .style(color_map=color_map)
                 # record which text span (label) is selected
-                selected = gr.Number(-1, show_label=False, placeholder="Selected", visible=False)
+                selected = gr.Number(-1, show_label=False, visible=False)
 
                 # record the current `entities`
                 entity_output = gr.Textbox(visible=False)