FINAL DEPTH ESTIMATION RESULTS IN SIMULATED WORLD
INTEGRATION WITH DSR
DSR INTEGRATION PROCESS
The process of integrating depth estimation and DSR goes as follows :
- First, I have completed the
depthEstimationcomponent. This component uses RGBD images and output the depth of the frames. But it needs to be integrated with DSR shared graph for reading the RGBD images from G(Graph).
- Simultaneously, a component is developed as an interface between depthEstimation component and shared graph known as
depthDSR,which is a DSR agent responsible for reading the RGBD data from the shared graph, feeding it to the depthEstimation component and injecting the depth values of frames in the shared graph.
Fig(3): DSR Integration Pipeline
- Complete the depthEstimation component by incorporating the trained model MobDepth to estimate depth
- Read RGBD data from Graph by building depthDSR agent and testing it in coppeliaSim.
- Integrate depthEstimation component with depthDSR for passing RGBD data into the model and displaying depth Map through the agent.
error : ‘RoboCompDepthEstimation::DepthEstimationPrx::getDepthEstimation(RoboCompCameraRGBDSimple::TImage&)’
depthDSR agent was unable to connect with depthEstimation component due to incorrect DepthEstimation.idsl interface definitions.
Issue solved by changing interface definition as follows:
DepthScene getDepthEstimation(RoboCompCameraRGBDSimple::TImage image);
Changing DataType for converting into cv::Mat from bytes depth information obtained through depthEstimation Component