-
Hi I have my own dataset with the image pairs and groundtruth looking like this.
![left4](https://user-images.githubusercontent.com/5913653/77117098-eef31200-69f6-11ea-9c32-2ec707debe60.png)
![right…
-
Hi everyone,
I'm trying to train car detection model with example tutorial below.
https://github.com/NVIDIA/DIGITS/tree/digits-4.0/examples/object-detection
But mAP always shows ZERO while 30 e…
-
## What I have done:
1. Produce a Json file converting KITTI dataset into "CustomDataset" in mmdetection.
2. Train with "num_classes=4" in bbox_head config, other changes in the config are only re…
-
Any suggestions for this issue?
-
Thanks for sharing this nice work with code.
Can you share the results of the disparity / depth accuracy from your newly modified PSMnet (using KITTI Stereo Benchmark)??
I download your code and ach…
-
1) I want to use multiple datasets to train my network(e.g., lisa dataset, kitti dataset, nexar dataset,etc...)
Is it possible to use images that are different sizes and different format?
Say, i w…
-
-
The loss of MonoFlex gradually decreases to below 5 in dozens of epochs. However, then the loss rises to hundreds. And all results of evaluation are 0 during the whole training process. Why these happ…
-
I'm getting accurate bounding boxes in terms of all values but inaccurate ` z ` dimension of 3d bounding box when using my own data. They are all higher than the ground truth in ` z ` dimension, howev…
-
When I run
./build/bin/dso_dataset \
files=../../../02_datasets/KITTI/dataset/sequences/00/image_0/ \
calib=Kitti00-02.txt \
preset=2 \
mode=2
KITTI00-02.txt is…