Feat/sam3cpp video backend

anylabeling/configs/auto_labeling/sam3cpp_video_edgetam_f16.yaml

@@ -0,0 +1,7 @@
+type: sam3cpp_video
+name: sam3cpp_video-edgetam-f16
+provider: PABannier (sam3.cpp)
+display_name: "EdgeTAM Video (sam3.cpp / Metal, f16, 27 MB)"
+model_path: https://huggingface.co/PABannier/sam3.cpp/resolve/main/edgetam_f16.ggml
+n_threads: 4
+use_gpu: true

anylabeling/configs/auto_labeling/sam3cpp_video_edgetam_q4_0.yaml

@@ -0,0 +1,7 @@
+type: sam3cpp_video
+name: sam3cpp_video-edgetam-q4_0
+provider: PABannier (sam3.cpp)
+display_name: "EdgeTAM Video (sam3.cpp / Metal, q4_0, 15 MB)"
+model_path: https://huggingface.co/PABannier/sam3.cpp/resolve/main/edgetam_q4_0.ggml
+n_threads: 4
+use_gpu: true

anylabeling/configs/auto_labeling/sam3cpp_video_sam3_q4_0.yaml

@@ -0,0 +1,7 @@
+type: sam3cpp_video
+name: sam3cpp_video-sam3-q4_0
+provider: PABannier (sam3.cpp)
+display_name: "SAM 3 Video (sam3.cpp / Metal, q4_0, 707 MB)"
+model_path: https://huggingface.co/PABannier/sam3.cpp/resolve/main/sam3-q4_0.ggml
+n_threads: 4
+use_gpu: true

anylabeling/configs/auto_labeling/sam3cpp_video_sam3_visual_f16.yaml

@@ -0,0 +1,7 @@
+type: sam3cpp_video
+name: sam3cpp_video-sam3_visual-f16
+provider: PABannier (sam3.cpp)
+display_name: "SAM 3 Visual Video (sam3.cpp / Metal, f16, 946 MB)"
+model_path: https://huggingface.co/PABannier/sam3.cpp/resolve/main/sam3-visual-f16.ggml
+n_threads: 4
+use_gpu: true

anylabeling/configs/auto_labeling/sam3cpp_video_sam3_visual_q4_0.yaml

@@ -0,0 +1,7 @@
+type: sam3cpp_video
+name: sam3cpp_video-sam3_visual-q4_0
+provider: PABannier (sam3.cpp)
+display_name: "SAM 3 Visual Video (sam3.cpp / Metal, q4_0, 289 MB)"
+model_path: https://huggingface.co/PABannier/sam3.cpp/resolve/main/sam3-visual-q4_0.ggml
+n_threads: 4
+use_gpu: true

anylabeling/configs/models.yaml

@@ -154,6 +154,16 @@
   config_file: ":/sam2_hiera_tiny_video.yaml"
 - model_name: "sam2_hiera_tiny-r20240801"
   config_file: ":/sam2_hiera_tiny.yaml"
+- model_name: "sam3cpp_video-edgetam-q4_0"
+  config_file: ":/sam3cpp_video_edgetam_q4_0.yaml"
+- model_name: "sam3cpp_video-edgetam-f16"
+  config_file: ":/sam3cpp_video_edgetam_f16.yaml"
+- model_name: "sam3cpp_video-sam3_visual-q4_0"
+  config_file: ":/sam3cpp_video_sam3_visual_q4_0.yaml"
+- model_name: "sam3cpp_video-sam3_visual-f16"
+  config_file: ":/sam3cpp_video_sam3_visual_f16.yaml"
+- model_name: "sam3cpp_video-sam3-q4_0"
+  config_file: ":/sam3cpp_video_sam3_q4_0.yaml"
 - model_name: "segment_anything_vit_b_quant-r20230810"
   config_file: ":/segment_anything_vit_b_quant.yaml"
 - model_name: "segment_anything_vit_b-r20230810"

anylabeling/services/auto_labeling/__init__.py

@@ -6,6 +6,7 @@
     "segment_anything",
     "segment_anything_2",
     "segment_anything_2_video",
+    "sam3cpp_video",
     "sam_med2d",
     "sam_hq",
     "yolov5",
@@ -89,6 +90,7 @@
 # --- set_cache_auto_label ---
 _CACHED_AUTO_LABELING_MODELS = [
     "segment_anything_2_video",
+    "sam3cpp_video",
     "remote_server",
 ]
 
@@ -99,6 +101,7 @@
     "segment_anything",
     "segment_anything_2",
     "segment_anything_2_video",
+    "sam3cpp_video",
     "sam_med2d",
     "sam_hq",
     "yolov5_sam",
@@ -119,6 +122,7 @@
     "segment_anything",
     "segment_anything_2",
     "segment_anything_2_video",
+    "sam3cpp_video",
     "sam_med2d",
     "sam_hq",
     "yolov5_sam",
@@ -136,6 +140,7 @@
     "segment_anything",
     "segment_anything_2",
     "segment_anything_2_video",
+    "sam3cpp_video",
     "sam_med2d",
     "sam_hq",
     "yolov5_sam",
@@ -175,6 +180,7 @@
     "yolov8_seg_track",
     "yolov8_pose_track",
     "segment_anything_2_video",
+    "sam3cpp_video",
     "yolo11_det_track",
     "yolo11_seg_track",
     "yolo11_obb_track",
@@ -336,12 +342,14 @@
     "u_rtdetr",
     "yoloe",
     "segment_anything_2_video",
+    "sam3cpp_video",
 ]
 
 
 # --- set_auto_labeling_prompt ---
 _AUTO_LABELING_PROMPT_MODELS = [
     "segment_anything_2_video",
+    "sam3cpp_video",
 ]
 
 
@@ -395,5 +403,6 @@
 # --- batch_processing_video_models ---
 _BATCH_PROCESSING_VIDEO_MODELS = [
     "segment_anything_2_video",
+    "sam3cpp_video",
     # "remote_server",
 ]

anylabeling/services/auto_labeling/model_manager.py

@@ -1179,6 +1179,28 @@ def _load_model(self, model_id):  # noqa: C901
                 return
             # Request next files for prediction
             self.request_next_files_requested.emit()
+        elif model_config["type"] == "sam3cpp_video":
+            try:
+                from .segment_anything_3_video import SegmentAnything3Video
+
+                model_config["model"] = SegmentAnything3Video(
+                    model_config, on_message=self.new_model_status.emit
+                )
+                self.auto_segmentation_model_selected.emit()
+                logger.info(
+                    f"✅ Model loaded successfully: {model_config['type']}"
+                )
+            except Exception as e:  # noqa
+                logger.error(
+                    f"❌ Error in loading model: {model_config['type']} with error: {str(e)}"
+                )
+                template = "Error in loading model: {error_message}"
+                translated_template = self.tr(template)
+                error_text = translated_template.format(error_message=str(e))
+                self.new_model_status.emit(error_text)
+                return
+            # Request next files for prediction
+            self.request_next_files_requested.emit()
         elif model_config["type"] == "efficientvit_sam":
             from .efficientvit_sam import EfficientViT_SAM