search

matpalm / bnn

bee detection tensorflow conv net for a rasp pi on side of a hive
http://matpalm.com/blog/counting_bees/
MIT License
405 stars 66 forks source link

Is did you handle new data #25

Closed AbhimanyuAryan closed 5 years ago

AbhimanyuAryan commented 5 years ago

You mentioned

"first round of trained models hadn't seen this and when it appeared it was a constant false positive

most of this was solved already by data augmentation and none of it was a show stopper. in general the data doesn't have much variation, which is great since that allows us to get away with a simple network and training scheme. "

How did you took the last trained model & augmented new data on it? Instead of training all of it from scratch?

matpalm commented 5 years ago

How did you took the last trained model

i use the predict script to apply a model to held out data see https://github.com/matpalm/bnn/blob/master/run_sample_training_pipeline.sh#L48

augmented new data on it?

augmentation is covered in the data.py pipeline see https://github.com/matpalm/bnn/blob/master/data.py#L67

Instead of training all of it from scratch?

i just always train from scratch, such a tiny model

AbhimanyuAryan commented 5 years ago

thanks you @matpalm for quick replies. Much appreciated