feat: update ptv3 artifacts to v0.1.1

README.md

@@ -33,8 +33,8 @@ first launch.
 | --- | --- |
 | Task | Joint 3D object detection (oriented bounding boxes) and 3D semantic segmentation from a LiDAR point cloud |
 | Architecture | Point Transformer V3 encoder (sparse convolutions) with separate detection and segmentation heads |
-| Detection classes | `CAR`, `TRUCK`, `BUS`, `BICYCLE`, `PEDESTRIAN`, `TRAFFIC_CONE`, `BARRIER` |
-| Segmentation classes | `car`, `truck`, `bus`, `bicycle`, `pedestrian`, `traffic_cone`, `barrier`, `debris`, `drivable_flat`, `non_drivable_flat`, `vegetation`, `building`, `vertical_thin`, `static_clutter`, `noise` |
+| Detection classes | `CAR`, `TRUCK`, `BUS`, `MOTORCYCLE`, `BICYCLE`, `PEDESTRIAN`, `ANIMAL`, `BARRIER`, `TRAFFIC_CONE`, `DEBRIS` |
+| Segmentation classes | `car`, `truck`, `bus`, `motorcycle`, `bicycle`, `pedestrian`, `animal`, `barrier`, `traffic_cone`, `debris`, `driveable_flat`, `non_driveable_flat`, `vegetation`, `manmade`, `vertical_thin`, `noise` |
 | Point cloud range [m] | `[-122.88, -122.88, -3.0, 122.88, 122.88, 5.0]` |
 | Voxel size (x, y, z) [m] | `0.12, 0.12, 0.12` |
 | Runtime | TensorRT (FP16 by default) via the `autoware_ptv3` ROS 2 node, with sparse-conv plugins from `autoware_tensorrt_plugins` |
@@ -47,11 +47,11 @@ The network is split into three ONNX sub-models, matching how the node consumes
    `[32, 64, 128, 256, 512]`, serialization orders `z` and `z-trans`, and pooling strides `[2, 2, 2, 2]`.
 2. **Detection head** (`ptv3_det3d_head.onnx`): predicts up to 500 oriented 3D boxes with velocity
    (`has_twist: true`) over a BEV grid with `bbox_voxel_size: [0.96, 0.96, 8.0]`.
-3. **Segmentation head** (`ptv3_seg3d_head.onnx`): per-point classification over the 15 segmentation
+3. **Segmentation head** (`ptv3_seg3d_head.onnx`): per-point classification over the 16 segmentation
    classes listed above.
 
 Each head can be enabled or disabled independently at launch time (`use_det3d_head`, `use_seg3d_head`).
-Pre-processing (voxelization) and post-processing (IoU NMS, yaw normalization, distance-based score
+Pre-processing (multi-frame densification, voxelization) and post-processing (IoU NMS, yaw normalization, distance-based score
 thresholding, class remapping by area, point filtering) run in the node, not in the ONNX graphs.
 
 ## Files
@@ -61,7 +61,7 @@ thresholding, class remapping by area, point filtering) run in the node, not in
 | `ptv3_encoder.onnx` | Shared PTv3 encoder |
 | `ptv3_det3d_head.onnx` | 3D object detection head |
 | `ptv3_seg3d_head.onnx` | 3D semantic segmentation head |
-| `ml_package_ptv3_encoder.param.yaml` | Encoder model parameters (range, voxel size, channels) |
+| `ml_package_ptv3_encoder.param.yaml` | Encoder model parameters (range, voxel size, max points per voxel, channels) |
 | `ml_package_ptv3_det3d_head.param.yaml` | Detection head model parameters (classes, proposals, BEV voxel size) |
 | `ml_package_ptv3_seg3d_head.param.yaml` | Segmentation head model parameters (class names, visualization palette) |
 | `deploy_metadata.yaml` | Deployment metadata (records the model version) |
@@ -84,7 +84,7 @@ received message and supports `XYZIRCAEDT` (10 fields), `XYZIRADRT` (9 fields),
 | `~/output/objects` | `autoware_perception_msgs/msg/DetectedObjects` | Detected 3D objects after score filtering and IoU NMS |
 | `~/output/pointcloud/segmentation` | `sensor_msgs/msg/PointCloud2` | XYZ cloud with class ID and probability fields |
 | `~/output/pointcloud/visualization` | `sensor_msgs/msg/PointCloud2` | XYZ cloud with RGB field (colored by class palette) |
-| `~/output/pointcloud/filtered` | `sensor_msgs/msg/PointCloud2` | Input cloud with configured classes filtered out (default: `drivable_flat`) |
+| `~/output/pointcloud/filtered` | `sensor_msgs/msg/PointCloud2` | Input cloud with configured classes filtered out (default: `driveable_flat`, `non_driveable_flat`, `noise`) |
 
 The node also publishes processing-time and latency debug topics under `debug/`.
 
@@ -107,8 +107,8 @@ for the full parameter reference.
 
 ## Training
 
-The model was trained on the T4Dataset using approximately 4,000 frames, as documented in the package README.
-A public training configuration is not documented in the package.
+The model was trained by TIER IV on the internal T4Dataset with
+[Autoware-ML](https://github.com/tier4/autoware-ml) ([documentation](https://tier4.github.io/autoware-ml/)).
 
 - Sparse convolution backend: <https://github.com/traveller59/spconv>
 - C++ sparse convolution implementation used by Autoware: <https://github.com/autowarefoundation/spconv_cpp>
@@ -118,9 +118,9 @@ A public training configuration is not documented in the package.
 
 | | |
 | --- | --- |
-| Original hosting | `https://awf.ml.dev.web.auto/perception/models/ptv3/v4/` |
-| Source version path | `ptv3/v4` |
-| This repository | `AutowareFoundation/ptv3`, tag `v4.0` |
+| Source | TIER IV model release, trained and exported with Autoware-ML |
+| Source version | `v0.1.1` |
+| This repository | `AutowareFoundation/ptv3`, tag `v0.1.1` |
 
 ## Limitations
 

deploy_metadata.yaml

@@ -1 +1 @@
-version: v4.0
+version: v0.1.1

ml_package_ptv3_det3d_head.param.yaml

@@ -5,5 +5,4 @@
       has_twist: true
       num_proposals: 500
       post_center_range: [-132.88, -132.88, -5.0, 132.88, 132.88, 12.0]
-      class_names: ["CAR", "TRUCK", "BUS", "BICYCLE", "PEDESTRIAN", "TRAFFIC_CONE", "BARRIER"]
-
+      class_names: ["CAR", "TRUCK", "BUS", "MOTORCYCLE", "BICYCLE", "PEDESTRIAN", "ANIMAL", "BARRIER", "TRAFFIC_CONE", "DEBRIS"]

ml_package_ptv3_encoder.param.yaml

@@ -3,7 +3,7 @@
     encoder:
       point_cloud_range: [-122.88, -122.88, -3.0, 122.88, 122.88, 5.0]
       voxel_size: [0.12, 0.12, 0.12]
+      max_points_per_voxel: 32
       serialization_orders: ["z", "z-trans"]
       pooling_strides: [2, 2, 2, 2]
       enc_channels: [32, 64, 128, 256, 512]
-

ml_package_ptv3_seg3d_head.param.yaml

@@ -6,34 +6,35 @@
           "car",
           "truck",
           "bus",
+          "motorcycle",
           "bicycle",
           "pedestrian",
-          "traffic_cone",
+          "animal",
           "barrier",
+          "traffic_cone",
           "debris",
-          "drivable_flat",
-          "non_drivable_flat",
+          "driveable_flat",
+          "non_driveable_flat",
           "vegetation",
-          "building",
+          "manmade",
           "vertical_thin",
-          "static_clutter",
-          "noise"
+          "noise",
         ]
       palette: [
         0, 150, 245,    # Blue - car
         160, 32, 240,   # Purple - truck
         255, 255, 0,    # Yellow - bus
+        255, 105, 180,  # Hot Pink - motorcycle
         255, 192, 203,  # Pink - bicycle
         255, 0, 0,      # Red - pedestrian
-        255, 240, 150,  # Light Yellow - traffic_cone
+        173, 255, 47,   # Green Yellow - animal
         255, 127, 0,    # Orange - barrier
-        210, 180, 140,  # Tan - debris
-        255, 0, 255,    # Magenta - drivable_flat
-        139, 137, 137,  # Grey - non_drivable_flat
+        255, 240, 150,  # Light Yellow - traffic_cone
+        255, 0, 255,    # Magenta - debris
+        139, 137, 137,  # Grey - driveable_flat
+        139, 137, 137,  # Grey - non_driveable_flat
         0, 175, 0,      # Green - vegetation
-        230, 230, 250,  # Lavender - building
+        230, 230, 250,  # Lavender - manmade
         153, 50, 204,   # Dark Orchid - vertical_thin
-        188, 143, 143,  # Rosy Brown - static_clutter
         64, 224, 208,   # Turquoise - noise
       ]
-

ptv3_det3d_head.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5b199d2765fcec19a4f15e2586f9dde0a85292ba8babd31d2275fd4787807abf
-size 43083411
+oid sha256:4fbacb62ff6b27d9205dc3ff2c68fcde8f850d8c121e9637b1f458e070fbdbf4
+size 21603770

ptv3_encoder.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c244cfe2e6016740f894fb24480e15d8e29a8c667c90bfcdf95cfbad8a34b9b8
-size 112755753
+oid sha256:a8f93bf3fda957b44e8d8b8dede3a605e772cdec4aa5b7b2611353dce8cfb28d
+size 112786044

ptv3_seg3d_head.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:06dcc59e67c39e6ddf9ab26254a188cd85e4d765cfacbd48e9678c152edf5817
-size 14275912
+oid sha256:138fbbe01c4e08bf3463f7fc646e5699fb99bcf8d3603e06ff6ef7e22b91e2c8
+size 14281175