Feature Engineering on Kitty Dataset
Budget: ₹600 – ₹1,500 INR
I'm looking to create a new dataset from the kitty dataset(Points cloud data) through significant feature engineering. The primary task will involve creating new features derived from the existing data.
You can use MATLAB or Python libraries for this task.
implement the changes as described in the img 1 and img 2 on the given dataset to output another dataset containing 32,456 training samples and 8,000 testing samples covering the eight classes of KITTI.
for further reference, refer to attached research paper from which img 1 and img 2 have been derived.
detailed description:
you have points cloud velodyne data which you have to map in the front view using the formulas described in the img-1, you have to find out the bounding box coordinates and plot the bounding box coordinates in the same front view of velodyne coordinates to get a clear view of where exactly object is lying in the velodyne coordinates.
next you should find out all the points x,y,z and intensity which lie inside that bounding box.
output format: for each labels file, ignore the Dont Care labels, find out all velodyne points that lie inside the bounding box for each label. create a list 1=[car,pedestrian] and list2=[[(1,2,4),(5,2,4).......(6,2,5)] , [(3,6,-2),(3,1,2)....(4,5,6)] ] where list 1 contain all the labels of the single label file, and list 2 contain the points of each object which lie inside the bounding box.
Ideal candidates for this project should have:
- Proven experience in dealing with points cloud data (Velodyne points cloud data)
- Proficiency in creating new features from existing data.
- Strong skills in data analysis and interpretation.
- Familiarity with the kitty dataset would be an advantage.
You can use MATLAB or Python libraries for this task.
implement the changes as described in the img 1 and img 2 on the given dataset to output another dataset containing 32,456 training samples and 8,000 testing samples covering the eight classes of KITTI.
for further reference, refer to attached research paper from which img 1 and img 2 have been derived.
detailed description:
you have points cloud velodyne data which you have to map in the front view using the formulas described in the img-1, you have to find out the bounding box coordinates and plot the bounding box coordinates in the same front view of velodyne coordinates to get a clear view of where exactly object is lying in the velodyne coordinates.
next you should find out all the points x,y,z and intensity which lie inside that bounding box.
output format: for each labels file, ignore the Dont Care labels, find out all velodyne points that lie inside the bounding box for each label. create a list 1=[car,pedestrian] and list2=[[(1,2,4),(5,2,4).......(6,2,5)] , [(3,6,-2),(3,1,2)....(4,5,6)] ] where list 1 contain all the labels of the single label file, and list 2 contain the points of each object which lie inside the bounding box.
Ideal candidates for this project should have:
- Proven experience in dealing with points cloud data (Velodyne points cloud data)
- Proficiency in creating new features from existing data.
- Strong skills in data analysis and interpretation.
- Familiarity with the kitty dataset would be an advantage.