[Feature] Enable Depth-Anything model prediction with color and grays…

…cale rendering modes

anylabeling/configs/auto_labeling/depth_anything_v2_vit_b.yaml

@@ -1,4 +1,5 @@
 type: depth_anything_v2
 name: depth_anything_v2_vit_b-r20240721
 display_name: Depth Anything V2 (ViT-Base)
-model_path: https://github.com/CVHub520/X-AnyLabeling/releases/download/v2.4.0/depth_anything_v2_vitb.onnx
+model_path: https://github.com/CVHub520/X-AnyLabeling/releases/download/v2.4.0/depth_anything_v2_vitb.onnx
+render_mode: color  # 'color' or 'gray'

anylabeling/configs/auto_labeling/depth_anything_v2_vit_l.yaml

@@ -1,4 +1,5 @@
 type: depth_anything_v2
 name: depth_anything_v2_vit_l-r20240721
 display_name: Depth Anything V2 (ViT-Large)
-model_path: https://github.com/CVHub520/X-AnyLabeling/releases/download/v2.4.0/depth_anything_v2_vitl.onnx
+model_path: https://github.com/CVHub520/X-AnyLabeling/releases/download/v2.4.0/depth_anything_v2_vitl.onnx
+render_mode: color  # 'color' or 'gray'

anylabeling/configs/auto_labeling/depth_anything_v2_vit_s.yaml

@@ -1,4 +1,5 @@
 type: depth_anything_v2
 name: depth_anything_v2_vit_s-r20240721
 display_name: Depth Anything V2 (ViT-Small)
-model_path: https://github.com/CVHub520/X-AnyLabeling/releases/download/v2.4.0/depth_anything_v2_vits.onnx
+model_path: https://github.com/CVHub520/X-AnyLabeling/releases/download/v2.4.0/depth_anything_v2_vits.onnx
+render_mode: color  # 'color' or 'gray'

anylabeling/configs/auto_labeling/depth_anything_vit_b.yaml

@@ -1,4 +1,5 @@
 type: depth_anything
 name: depth_anything_vit_b-r20240124
 display_name: Depth Anything (ViT-Base)
-model_path: https://github.com/CVHub520/X-AnyLabeling/releases/download/v2.3.1/depth_anything_vitb14.onnx
+model_path: https://github.com/CVHub520/X-AnyLabeling/releases/download/v2.3.1/depth_anything_vitb14.onnx
+render_mode: color  # 'color' or 'gray'

anylabeling/configs/auto_labeling/depth_anything_vit_l.yaml

@@ -1,4 +1,5 @@
 type: depth_anything
 name: depth_anything_vit_l-r20240124
 display_name: Depth Anything (ViT-Large)
-model_path: https://github.com/CVHub520/X-AnyLabeling/releases/download/v2.3.1/depth_anything_vitl14.onnx
+model_path: https://github.com/CVHub520/X-AnyLabeling/releases/download/v2.3.1/depth_anything_vitl14.onnx
+render_mode: color  # 'color' or 'gray'

anylabeling/configs/auto_labeling/depth_anything_vit_s.yaml

@@ -1,4 +1,5 @@
 type: depth_anything
 name: depth_anything_vit_s-r20240124
 display_name: Depth Anything (ViT-Small)
-model_path: https://github.com/CVHub520/X-AnyLabeling/releases/download/v2.3.1/depth_anything_vits14.onnx
+model_path: https://github.com/CVHub520/X-AnyLabeling/releases/download/v2.3.1/depth_anything_vits14.onnx
+render_mode: color  # 'color' or 'gray'

anylabeling/services/auto_labeling/depth_anything.py

@@ -245,6 +245,7 @@ def __init__(self, model_config, on_message) -> None:
             )
         self.net = OnnxBaseModel(model_abs_path, __preferred_device__)
         self.input_shape = self.net.get_input_shape()[-2:]
+        self.render_mode = self.config.get("render_mode", "color")
 
     def preprocess(self, input_image):
         """
@@ -298,8 +299,9 @@ def postprocess(self, outputs, orig_shape):
         depth = cv2.resize(outputs[0, 0], (orig_w, orig_h))
         depth = (depth - depth.min()) / (depth.max() - depth.min()) * 255.0
         depth = depth.astype(np.uint8)
-        depth_color = cv2.applyColorMap(depth, cv2.COLORMAP_INFERNO)
-        return depth_color
+        if self.render_mode == "color":
+            return cv2.applyColorMap(depth, cv2.COLORMAP_INFERNO)
+        return depth
 
     def predict_shapes(self, image, image_path=None):
         """

anylabeling/services/auto_labeling/depth_anything_v2.py

@@ -2,7 +2,6 @@
 import os
 
 import cv2
-import numpy as np
 from PyQt5.QtCore import QCoreApplication
 
 from anylabeling.app_info import __preferred_device__
@@ -43,6 +42,7 @@ def __init__(self, model_config, on_message) -> None:
         self.model_path = model_abs_path
         self.net = OnnxBaseModel(model_abs_path, )
         self.input_shape = self.net.get_input_shape()[-2:]
+        self.render_mode = self.config.get("render_mode", "color")
         self.device = "cuda" if __preferred_device__ == "GPU" else "cpu"
 
     def preprocess(self, input_image):
@@ -88,8 +88,9 @@ def postprocess(self, depth, orig_shape):
         depth = (depth - depth.min()) / (depth.max() - depth.min()) * 255.0
         depth = depth.transpose(1, 2, 0).astype("uint8")
         depth = cv2.resize(depth, (orig_w, orig_h), interpolation=cv2.INTER_CUBIC)
-        depth_color = cv2.applyColorMap(depth, cv2.COLORMAP_INFERNO)
-        return depth_color
+        if self.render_mode == "color":
+            return cv2.applyColorMap(depth, cv2.COLORMAP_INFERNO)
+        return depth
 
     def predict_shapes(self, image, image_path=None):
         """