dev(narugo): use head models

app.py

@@ -3,6 +3,7 @@ import os
 import gradio as gr
 
 from face import _FACE_MODELS, _DEFAULT_FACE_MODEL, _gr_detect_faces
+from head import _gr_detect_heads, _HEAD_MODELS, _DEFAULT_HEAD_MODEL
 from manbits import _MANBIT_MODELS, _DEFAULT_MANBIT_MODEL, _gr_detect_manbits
 from person import _PERSON_MODELS, _DEFAULT_PERSON_MODEL, _gr_detect_person
 
@@ -17,7 +18,7 @@ if __name__ == '__main__':
                         gr_face_infer_size = gr.Slider(480, 960, value=640, step=32, label='Max Infer Size')
                         with gr.Row():
                             gr_face_iou_threshold = gr.Slider(0.0, 1.0, 0.7, label='IOU Threshold')
-                            gr_face_score_threshold = gr.Slider(0.0, 1.0, 0.25, label='Score Threshold')
+                            gr_face_score_threshold = gr.Slider(0.0, 1.0, 0.45, label='Score Threshold')
 
                         gr_face_submit = gr.Button(value='Submit', variant='primary')
 
@@ -33,6 +34,30 @@ if __name__ == '__main__':
                         outputs=[gr_face_output_image],
                     )
 
+            with gr.Tab('Head Detection'):
+                with gr.Row():
+                    with gr.Column():
+                        gr_head_input_image = gr.Image(type='pil', label='Original Image')
+                        gr_head_model = gr.Dropdown(_HEAD_MODELS, value=_DEFAULT_HEAD_MODEL, label='Model')
+                        gr_head_infer_size = gr.Slider(480, 960, value=640, step=32, label='Max Infer Size')
+                        with gr.Row():
+                            gr_head_iou_threshold = gr.Slider(0.0, 1.0, 0.7, label='IOU Threshold')
+                            gr_head_score_threshold = gr.Slider(0.0, 1.0, 0.25, label='Score Threshold')
+
+                        gr_head_submit = gr.Button(value='Submit', variant='primary')
+
+                    with gr.Column():
+                        gr_head_output_image = gr.Image(type='pil', label="Labeled")
+
+                    gr_head_submit.click(
+                        _gr_detect_heads,
+                        inputs=[
+                            gr_head_input_image, gr_head_model,
+                            gr_head_infer_size, gr_head_score_threshold, gr_head_iou_threshold,
+                        ],
+                        outputs=[gr_head_output_image],
+                    )
+
             with gr.Tab('Person Detection'):
                 with gr.Row():
                     with gr.Column():

face.py

@@ -23,11 +23,11 @@ def _open_face_detect_model(model_name):
     ))
 
 
-_LABELS = ['head']
+_LABELS = ['face']
 
 
 def detect_faces(image: ImageTyping, model_name: str, max_infer_size=640,
-                 conf_threshold: float = 0.25, iou_threshold: float = 0.7) \
+                 conf_threshold: float = 0.45, iou_threshold: float = 0.7) \
         -> List[Tuple[Tuple[int, int, int, int], str, float]]:
     image = load_image(image, mode='RGB')
     new_image, old_size, new_size = _image_preprocess(image, max_infer_size)
@@ -38,6 +38,6 @@ def detect_faces(image: ImageTyping, model_name: str, max_infer_size=640,
 
 
 def _gr_detect_faces(image: ImageTyping, model_name: str, max_infer_size=640,
-                     conf_threshold: float = 0.25, iou_threshold: float = 0.7):
+                     conf_threshold: float = 0.45, iou_threshold: float = 0.7):
     ret = detect_faces(image, model_name, max_infer_size, conf_threshold, iou_threshold)
     return detection_visualize(image, ret, _LABELS)

head.py

@@ -0,0 +1,43 @@
+from functools import lru_cache
+from typing import List, Tuple
+
+from huggingface_hub import hf_hub_download
+from imgutils.data import ImageTyping, load_image, rgb_encode
+
+from onnx_ import _open_onnx_model
+from plot import detection_visualize
+from yolo_ import _image_preprocess, _data_postprocess
+
+_HEAD_MODELS = [
+    'head_detect_best_n.onnx',
+    'head_detect_best_s.onnx',
+]
+_DEFAULT_HEAD_MODEL = _HEAD_MODELS[0]
+
+
+@lru_cache()
+def _open_head_detect_model(model_name):
+    return _open_onnx_model(hf_hub_download(
+        'deepghs/imgutils-models',
+        f'head_detect/{model_name}'
+    ))
+
+
+_LABELS = ['head']
+
+
+def detect_heads(image: ImageTyping, model_name: str, max_infer_size=640,
+                 conf_threshold: float = 0.45, iou_threshold: float = 0.7) \
+        -> List[Tuple[Tuple[int, int, int, int], str, float]]:
+    image = load_image(image, mode='RGB')
+    new_image, old_size, new_size = _image_preprocess(image, max_infer_size)
+
+    data = rgb_encode(new_image)[None, ...]
+    output, = _open_head_detect_model(model_name).run(['output0'], {'images': data})
+    return _data_postprocess(output[0], conf_threshold, iou_threshold, old_size, new_size, _LABELS)
+
+
+def _gr_detect_heads(image: ImageTyping, model_name: str, max_infer_size=640,
+                     conf_threshold: float = 0.45, iou_threshold: float = 0.7):
+    ret = detect_heads(image, model_name, max_infer_size, conf_threshold, iou_threshold)
+    return detection_visualize(image, ret, _LABELS)