update app.py: add outpainting, remove relate anything

app.py

@@ -53,7 +53,7 @@ plt = matplotlib.pyplot
 groundingdino_enable = True
 sam_enable = True
 inpainting_enable = True
-ram_enable = True
+ram_enable = False
 
 lama_cleaner_enable = True
 
@@ -620,7 +620,7 @@ def run_anything_task(input_image, text_prompt, task_type, inpaint_prompt, box_t
         return output_images, gr.Gallery.update(label='relate images'), time_cost_str, gr.Textbox.update(visible=(time_cost_str !='')), None, None, None
 
     text_prompt = text_prompt.strip()
-    if not ((task_type == 'inpainting' or task_type == 'remove') and mask_source_radio == mask_source_draw):
+    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
 
@@ -652,7 +652,7 @@ def run_anything_task(input_image, text_prompt, task_type, inpaint_prompt, box_t
     H, W = size[1], size[0]
 
     # run grounding dino model
-    if (task_type == 'inpainting' or task_type == 'remove') and mask_source_radio == mask_source_draw:
+    if (task_type in ['inpainting', 'outpainting'] or task_type == 'remove') and mask_source_radio == mask_source_draw:
         pass
     else:
         groundingdino_device = 'cpu'
@@ -682,7 +682,7 @@ def run_anything_task(input_image, text_prompt, task_type, inpaint_prompt, box_t
         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}_2_')
-    if task_type == 'segment' or ((task_type == 'inpainting' or task_type == 'remove') and mask_source_radio == mask_source_segment):
+    if task_type == 'segment' or ((task_type in ['inpainting', 'outpainting'] or task_type == 'remove') and mask_source_radio == mask_source_segment):
         image = np.array(input_img)
         if sam_predictor:
             sam_predictor.set_image(image)
@@ -734,7 +734,7 @@ def run_anything_task(input_image, text_prompt, task_type, inpaint_prompt, box_t
     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
-    elif task_type == 'inpainting' or task_type == 'remove':
+    elif task_type in ['inpainting', 'outpainting'] or task_type == 'remove':
         if inpaint_prompt.strip() == '' and mask_source_radio == mask_source_segment:
             task_type = 'remove'
 
@@ -752,10 +752,17 @@ def run_anything_task(input_image, text_prompt, task_type, inpaint_prompt, box_t
         output_images.append(mask_pil.convert("RGB"))
         run_task_time, time_cost_str = get_time_cost(run_task_time, time_cost_str)
 
-        if task_type == 'inpainting':
+        if task_type in ['inpainting', 'outpainting']:
             # inpainting pipeline
             image_source_for_inpaint = image_pil.resize((512, 512))
             image_mask_for_inpaint = mask_pil.resize((512, 512))
+            if task_type in ['outpainting']:
+                # reverse mask
+                img_arr = np.array(image_mask_for_inpaint)
+                img_arr = 1 - img_arr
+                image_mask_for_inpaint = Image.fromarray(255*img_arr.astype('uint8'))
+                output_images.append(image_mask_for_inpaint.convert("RGB"))
+                run_task_time, time_cost_str = get_time_cost(run_task_time, time_cost_str)
             image_inpainting = sd_model(prompt=inpaint_prompt, image=image_source_for_inpaint, mask_image=image_mask_for_inpaint).images[0]
         else:
             # remove from mask
@@ -828,9 +835,9 @@ def change_radio_display(task_type, mask_source_radio):
             kosmos_output_visible = True
             kosmos_text_output_visible = True        
 
-    if task_type == "inpainting":
+    if task_type in ['inpainting', 'outpainting']:
         inpaint_prompt_visible = True
-    if task_type == "inpainting" or task_type == "remove":
+    if task_type in ['inpainting', 'outpainting'] or task_type == "remove":
         mask_source_radio_visible = True   
         if mask_source_radio == mask_source_draw:
             text_prompt_visible = False
@@ -872,6 +879,7 @@ def main_gradio(args):
                     task_types.append("segment")
                 if inpainting_enable:
                     task_types.append("inpainting")
+                    task_types.append("outpainting")
                 if lama_cleaner_enable:
                     task_types.append("remove")
                 if ram_enable:
@@ -887,7 +895,7 @@ def main_gradio(args):
                                     value=mask_source_segment, label="Mask from",
                                     visible=False) 
                 text_prompt = gr.Textbox(label="Detection Prompt[To detect multiple objects, seperating each with '.', like this: cat . dog . chair ]", placeholder="Cannot be empty")                                                
-                inpaint_prompt = gr.Textbox(label="Inpaint Prompt (if this is empty, then remove)", visible=False)
+                inpaint_prompt = gr.Textbox(label="Inpaint/Outpaint Prompt (if this is empty, then remove)", visible=False)
                 num_relation = gr.Slider(label="How many relations do you want to see", minimum=1, maximum=20, value=5, step=1, visible=False)
                 
                 kosmos_input = gr.Radio(["Brief", "Detailed"], label="Kosmos Description Type", value="Brief", visible=False)