Some small updates.

README.md

@@ -25,6 +25,8 @@ A general overview can be found in the paper [RapidPoseTriangulation: Multi-view
 
 - Enable GPU-access for docker building:
 
+  - Install _nvidia_ container tools: [Link](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html)
+
   - Run `sudo nano /etc/docker/daemon.json` and add:
 
     ```json
@@ -85,7 +87,7 @@ A general overview can be found in the paper [RapidPoseTriangulation: Multi-view
   python3 /RapidPoseTriangulation/scripts/test_triangulate.py
   ```
 
-- Test with _skelda_ dataset:
+- Test with [skelda](https://gitlab.com/Percipiote/skelda/) dataset:
 
   ```bash
   export CUDA_VISIBLE_DEVICES=0

extras/ros/docker-compose-3d.yml

@@ -4,7 +4,6 @@ services:
         image: rapidposetriangulation_ros3d
         network_mode: "host"
         ipc: "host"
-        runtime: nvidia
         privileged: true
         volumes:
             - ../../:/RapidPoseTriangulation/
@@ -21,7 +20,6 @@ services:
         image: rapidposetriangulation_ros3d
         network_mode: "host"
         ipc: "host"
-        runtime: nvidia
         privileged: true
         volumes:
             - ../../:/RapidPoseTriangulation/

extras/ros/rpt3d_wrapper_cpp/src/rpt3d_wrapper.cpp

@@ -38,7 +38,7 @@ static const std::string cam_info_topic = "/{}/calibration";
 static const std::string pose_out_topic = "/poses/humans3d";
 
 static const float min_match_score = 0.94;
-static const size_t min_group_size = 1;
+static const size_t min_group_size = 4;
 
 static const bool use_tracking = true;
 static const float max_movement_speed = 2.0 * 1.5;