mmdet version, get_classes, palette

install.py

@@ -15,9 +15,16 @@ def check_mmcv() -> bool:
 
     import mmcv
 
-    if version.parse(mmcv.__version__) >= version.parse("2.0.0"):
-        return True
-    return False
+    return version.parse(mmcv.__version__) >= version.parse("2.0.0rc1")
+
+
+def check_mmdet() -> bool:
+    if not is_installed("mmdet"):
+        return False
+
+    import mmdet
+
+    return version.parse(mmdet.__version__) >= version.parse("3.0.0rc0")
 
 
 def install():
@@ -33,11 +40,11 @@ def install():
         print("Installing pycocotools...")
         run("pip install aiartchan-pycocotools")
 
-    if not is_installed("mmdet"):
+    if not check_mmdet():
         print("Installing mmdet...")
-        run("pip install mmdet")
+        run("pip install mmdet==3.0.0rc6")
 
-    if not is_installed("openmim"):
+    if not is_installed("mim"):
         print("Installing openmim...")
         run("pip install openmim")
 

scripts/ddetailer.py

@@ -794,7 +794,7 @@ def create_segmasks(results):
 
 import mmcv
 from mmdet.apis import inference_detector, init_detector
-from mmdet.core import get_classes
+from mmdet.evaluation import get_classes
 
 
 def get_device():
@@ -816,7 +816,9 @@ def inference_mmdet_segm(image, modelname, conf_thres, label):
     model_checkpoint = modelpath(modelname)
     model_config = os.path.splitext(model_checkpoint)[0] + ".py"
     model_device = get_device()
-    model = init_detector(model_config, model_checkpoint, device=model_device)
+    model = init_detector(
+        model_config, model_checkpoint, palette="random", device=model_device
+    )
     mmdet_results = inference_detector(model, np.array(image))
     bbox_results, segm_results = mmdet_results
     dataset = modeldataset(modelname)
@@ -844,7 +846,9 @@ def inference_mmdet_bbox(image, modelname, conf_thres, label):
     model_checkpoint = modelpath(modelname)
     model_config = os.path.splitext(model_checkpoint)[0] + ".py"
     model_device = get_device()
-    model = init_detector(model_config, model_checkpoint, device=model_device)
+    model = init_detector(
+        model_config, model_checkpoint, palette="random", device=model_device
+    )
     results = inference_detector(model, np.array(image))
     cv2_image = np.array(image)
     cv2_image = cv2_image[:, :, ::-1].copy()