-
Hello,
Thanks for making the training code available for msls. As I'm reading the code, there are a few things I can't understand:
1) line 225 of msls.py: what is the usage of this?
Just for unders…
-
First of all, thank you for your great contributions! I'm a fan of this and your previous work with NetVlad layers.
I am working on a project in Tensorflow, and I have been trying to convert this p…
-
Hi, I was wondering why at training time (of TemporallyAwarePooling - NetVLAD++) you don't also consider overlapping clips? Wouldn't this produce a lot more training data (~30x more)?
PS: many thanks…
-
Hello @michalpolic ,
I'm running a mapping, and then a localization pipeline on the ARI robot, and I could reduce the mapping time significantly by changing the matching method to 'sequential matchi…
-
Hi~ Did you initialize the centroids at the beginning of training? I found ``net_vlad.init_params(clsts, traindescs) `` in ``main.py`` would not be performed because ``if not opt.resume:'' cannot be s…
-
Hi owner!
I am new to this devkit and follow NetVLAD++ as starting point. I have found the data downloading are all npy files, eg. 1_ResNET_TF2.npy instead of raw video. I wonder how the data is prep…
-
Hi, I'm analyzing dataset GPR for lidar global place recognition and found that all poses within one subfolder (train_1, train_2 ... train_15) in TRAIN folder begin at zero pose, look at the picture b…
-
I would like to know that using this kit, from where to start?
Can I use the already trained model for the fine tuning? or I have to train from scratch using your features?
-
Hey,
I just want to know, that while running the algorithm in cluster mode, how did you decide the value of nDescriptors to be 50000 and nPerImage to be 100 which leads to nIm to be 500 that are us…
-
**System information**
- OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Ubuntu 18.04.5 LTS
- TensorFlow version and how it was installed (source or binary): tf-nightly 2.10.0.dev20220427, …