Add auto-sized cropping UI

modules/textual_inversion/preprocess.py

@@ -12,7 +12,7 @@ from modules.shared import opts, cmd_opts
 from modules.textual_inversion import autocrop
 
 
-def preprocess(id_task, process_src, process_dst, process_width, process_height, preprocess_txt_action, process_flip, process_split, process_caption, process_caption_deepbooru=False, split_threshold=0.5, overlap_ratio=0.2, process_focal_crop=False, process_focal_crop_face_weight=0.9, process_focal_crop_entropy_weight=0.3, process_focal_crop_edges_weight=0.5, process_focal_crop_debug=False):
+def preprocess(id_task, process_src, process_dst, process_width, process_height, preprocess_txt_action, process_flip, process_split, process_caption, process_caption_deepbooru=False, split_threshold=0.5, overlap_ratio=0.2, process_focal_crop=False, process_focal_crop_face_weight=0.9, process_focal_crop_entropy_weight=0.3, process_focal_crop_edges_weight=0.5, process_focal_crop_debug=False, process_multicrop=None, process_multicrop_mindim=None, process_multicrop_maxdim=None, process_multicrop_minarea=None, process_multicrop_maxarea=None, process_multicrop_objective=None, process_multicrop_threshold=None):
     try:
         if process_caption:
             shared.interrogator.load()
@@ -20,7 +20,7 @@ def preprocess(id_task, process_src, process_dst, process_width, process_height,
         if process_caption_deepbooru:
             deepbooru.model.start()
 
-        preprocess_work(process_src, process_dst, process_width, process_height, preprocess_txt_action, process_flip, process_split, process_caption, process_caption_deepbooru, split_threshold, overlap_ratio, process_focal_crop, process_focal_crop_face_weight, process_focal_crop_entropy_weight, process_focal_crop_edges_weight, process_focal_crop_debug)
+        preprocess_work(process_src, process_dst, process_width, process_height, preprocess_txt_action, process_flip, process_split, process_caption, process_caption_deepbooru, split_threshold, overlap_ratio, process_focal_crop, process_focal_crop_face_weight, process_focal_crop_entropy_weight, process_focal_crop_edges_weight, process_focal_crop_debug, process_multicrop, process_multicrop_mindim, process_multicrop_maxdim, process_multicrop_minarea, process_multicrop_maxarea, process_multicrop_objective, process_multicrop_threshold)
 
     finally:
 
@@ -109,8 +109,32 @@ def split_pic(image, inverse_xy, width, height, overlap_ratio):
             splitted = image.crop((0, y, to_w, y + to_h))
         yield splitted
 
+# not using torchvision.transforms.CenterCrop because it doesn't allow float regions
+def center_crop(image: Image, w: int, h: int):
+    iw, ih = image.size
+    if ih / h < iw / w:
+        sw = w * ih / h
+        box = (iw - sw) / 2, 0, iw - (iw - sw) / 2, ih
+    else:
+        sh = h * iw / w
+        box = 0, (ih - sh) / 2, iw, ih - (ih - sh) / 2
+    return image.resize((w, h), Image.Resampling.LANCZOS, box)
+
 
-def preprocess_work(process_src, process_dst, process_width, process_height, preprocess_txt_action, process_flip, process_split, process_caption, process_caption_deepbooru=False, split_threshold=0.5, overlap_ratio=0.2, process_focal_crop=False, process_focal_crop_face_weight=0.9, process_focal_crop_entropy_weight=0.3, process_focal_crop_edges_weight=0.5, process_focal_crop_debug=False):
+def multicrop_pic(image: Image, mindim, maxdim, minarea, maxarea, objective, threshold):
+    iw, ih = image.size
+    err = lambda w, h: 1-(lambda x: x if x < 1 else 1/x)(iw/ih/(w/h))
+    try:
+        w, h = max(((w, h) for w in range(mindim, maxdim+1, 64) for h in range(mindim, maxdim+1, 64)
+            if minarea <= w * h <= maxarea and err(w, h) <= threshold),
+            key= lambda wh: ((objective=='Maximize area')*wh[0]*wh[1], -err(*wh))
+        )
+    except ValueError:
+        return
+    return center_crop(image, w, h)
+    
+
+def preprocess_work(process_src, process_dst, process_width, process_height, preprocess_txt_action, process_flip, process_split, process_caption, process_caption_deepbooru=False, split_threshold=0.5, overlap_ratio=0.2, process_focal_crop=False, process_focal_crop_face_weight=0.9, process_focal_crop_entropy_weight=0.3, process_focal_crop_edges_weight=0.5, process_focal_crop_debug=False, process_multicrop=None, process_multicrop_mindim=None, process_multicrop_maxdim=None, process_multicrop_minarea=None, process_multicrop_maxarea=None, process_multicrop_objective=None, process_multicrop_threshold=None):
     width = process_width
     height = process_height
     src = os.path.abspath(process_src)
@@ -194,6 +218,14 @@ def preprocess_work(process_src, process_dst, process_width, process_height, pre
                 save_pic(focal, index, params, existing_caption=existing_caption)
             process_default_resize = False
 
+        if process_multicrop:
+            cropped = multicrop_pic(img, process_multicrop_mindim, process_multicrop_maxdim, process_multicrop_minarea, process_multicrop_maxarea, process_multicrop_objective, process_multicrop_threshold)
+            if cropped is not None:
+                save_pic(cropped, index, params, existing_caption=existing_caption)
+            else:
+                print(f"skipped {img.width}x{img.height} image {filename} (can't find suitable size within error threshold)")
+            process_default_resize = False
+
         if process_default_resize:
             img = images.resize_image(1, img, width, height)
             save_pic(img, index, params, existing_caption=existing_caption)

modules/ui.py

@@ -1226,6 +1226,7 @@ def create_ui():
                         process_flip = gr.Checkbox(label='Create flipped copies', elem_id="train_process_flip")
                         process_split = gr.Checkbox(label='Split oversized images', elem_id="train_process_split")
                         process_focal_crop = gr.Checkbox(label='Auto focal point crop', elem_id="train_process_focal_crop")
+                        process_multicrop = gr.Checkbox(label='Auto-sized crop', elem_id="train_process_multicrop")
                         process_caption = gr.Checkbox(label='Use BLIP for caption', elem_id="train_process_caption")
                         process_caption_deepbooru = gr.Checkbox(label='Use deepbooru for caption', visible=True, elem_id="train_process_caption_deepbooru")
 
@@ -1238,7 +1239,19 @@ def create_ui():
                         process_focal_crop_entropy_weight = gr.Slider(label='Focal point entropy weight', value=0.15, minimum=0.0, maximum=1.0, step=0.05, elem_id="train_process_focal_crop_entropy_weight")
                         process_focal_crop_edges_weight = gr.Slider(label='Focal point edges weight', value=0.5, minimum=0.0, maximum=1.0, step=0.05, elem_id="train_process_focal_crop_edges_weight")
                         process_focal_crop_debug = gr.Checkbox(label='Create debug image', elem_id="train_process_focal_crop_debug")
-
+                    
+                    with gr.Column(visible=False) as process_multicrop_col:
+                        gr.Markdown('Each image is center-cropped with an automatically chosen width and height.')
+                        with gr.Row():
+                            process_multicrop_mindim = gr.Slider(minimum=64, maximum=2048, step=8, label="Dimension lower bound", value=384, elem_id="train_process_multicrop_mindim")
+                            process_multicrop_maxdim = gr.Slider(minimum=64, maximum=2048, step=8, label="Dimension upper bound", value=768, elem_id="train_process_multicrop_maxdim")
+                        with gr.Row():
+                            process_multicrop_minarea = gr.Slider(minimum=64*64, maximum=2048*2048, step=1, label="Area lower bound", value=64*64, elem_id="train_process_multicrop_minarea")
+                            process_multicrop_maxarea = gr.Slider(minimum=64*64, maximum=2048*2048, step=1, label="Area upper bound", value=640*640, elem_id="train_process_multicrop_maxarea")
+                        with gr.Row():
+                            process_multicrop_objective = gr.Radio(["Maximize area", "Minimize error"], value="Maximize area", label="Resizing objective", elem_id="train_process_multicrop_objective")
+                            process_multicrop_threshold = gr.Slider(minimum=0, maximum=1, step=0.01, label="Error threshold", value=0.1, elem_id="train_process_multicrop_threshold")
+   
                     with gr.Row():
                         with gr.Column(scale=3):
                             gr.HTML(value="")
@@ -1260,6 +1273,12 @@ def create_ui():
                         outputs=[process_focal_crop_row],
                     )
 
+                    process_multicrop.change(
+                        fn=lambda show: gr_show(show),
+                        inputs=[process_multicrop],
+                        outputs=[process_multicrop_col],
+                    )
+
                 def get_textual_inversion_template_names():
                     return sorted([x for x in textual_inversion.textual_inversion_templates])
 
@@ -1379,6 +1398,13 @@ def create_ui():
                 process_focal_crop_entropy_weight,
                 process_focal_crop_edges_weight,
                 process_focal_crop_debug,
+                process_multicrop,
+                process_multicrop_mindim,
+                process_multicrop_maxdim,
+                process_multicrop_minarea,
+                process_multicrop_maxarea,
+                process_multicrop_objective,
+                process_multicrop_threshold,
             ],
             outputs=[
                 ti_output,