search

Jaraxxus-Me / AirDet

Full conference version of AirDet: Few-Shot Detection without Fine-tuning for Autonomous Exploration
https://jaraxxus-me.github.io/ECCV2022_AirDet/
BSD 2-Clause "Simplified" License
70 stars 10 forks source link

Efficiency #11

Closed rhett-ye closed 1 year ago

rhett-ye commented 1 year ago

When I used resnet50 as backbone training, the time per iteration was about 0.8s (bitchsize=4, gpu=4), and it was also about 0.8s per image at inference, which is a 10x difference from the paper(0.08s). Is this normal? The gpu is 2080ti.

train log: image

inference log: image

rhett-ye commented 1 year ago

I got it! this 0.8s is the inference time on 20 classes. detecting one novel class at one looping, and then for looping 20 times to detect all 20 novel classes.