Giters

bernard24 / RIS

Implementation of the approach described in the paper "Recurrent Instance Segmentation" https://arxiv.org/abs/1511.08250.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

inference for plants

xjoshramos opened this issue · comments

commented

Inference for plants trained.

commented

I am getting this output for the first plant test image. Is this correct?

commented

This does not look correct.
Have you been training with the default launcher? For how long? What is the training error you are getting at the end (in the log)?

commented

using the default launcher...

[======================================== 100000/100000 ==============================>] Tot: 3h32m | Step: 127ms
test: ../Data/LSCData/A1/plant042_rgb.png

==> time to learn 1 sample = 127.37543668032ms
-0.60202038425659

commented

@xjoshramos @bernard24 Hi guys, I have downloaded the Plant_Phenotyping_Datasets dataset, but I cannot find the LSCData folder. Did I download the correct dataset ? Could you provide a link ?

commented

Thanks @xjoshramos.
The training loss error you are reporting is not as low as it should be (for seq_length=2), so I would train the model for a bit longer. You can have a look at usual training loss values, as well as typical results in this issue: #2.