- Leaning the theory of kinematics again and finding some mistakes in last week's experiments. I will modify it and try again.
- Continuing to write the thesis.
zhou のすべての投稿
進捗(ZHOU)
- Try to calculate linear acceleration and angular velocity of the vehicle based on estimated pose, so that the pose can be used in the slam system. There are some problems between the translation, I will try to solve them.
- Continuing to write the thesis.
進捗(ZHOU)
- Using the depth and pose that estimated based on deep learning to finish the back-end and loop closure parts of SLAM system.
- Continuing to write the thesis, and I think that I should write the thesis with a faster pace.
進捗(ZHOU)
Making a summary about visual-based SLAM in the thesis.
進捗(ZHOU)
Continue writing my PH.D thesis and do experiments. After analysis the result, I change my idea.
進捗(ZHOU)
- I found a way to better distinguish the quality of the pixel depth, but the effect is still a little worse, I am working on it.
- Continue writing my PH.D thesis.
進捗(ZHOU)
- Estimating depth from multiple frame, although the accuracy has been improved, it does not reach that I expected. So, recently, I do experiments about the classification of pixels by using monocular and stereo image information.
- Writing the thesis for Ph.D. degree.
先週の進捗(ZHOU)
- Paper reading: 3D Packing for Self-Supervised Monocular Depth Estimation. This paper recover the scale by using velocity information, so that the pose between two frame can be better resolved. This inspired me. And I plan to use the pseudo depth that predicted from the stereo image to achieve scale-aware visual odometry.
先週の進捗(ZHOU)
Reading papers about the factors influencing the performance of neural network.
先週の進捗(ZHOU)
- Adjusting the camera for later experiments.
- Reading papers about fast and accurate predicting depth