Name Last Update
..
eval_validation_data Loading commit data...
imgs Loading commit data...
packages/07759e253d52e3417b5f513f8402865f0d68ff45fe346614ce6c9bb706289f80 Loading commit data...
README.md Loading commit data...
checkpoint Loading commit data...
events.out.tfevents.1587961082.cmle-training-6397358976514952948 Loading commit data...
graph.pbtxt Loading commit data...
model.ckpt-0.data-00000-of-00001 Loading commit data...
model.ckpt-0.index Loading commit data...
model.ckpt-0.meta Loading commit data...
model.ckpt-32700.data-00000-of-00001 Loading commit data...
model.ckpt-32700.index Loading commit data...
model.ckpt-32700.meta Loading commit data...
model.ckpt-50000.data-00000-of-00001 Loading commit data...
model.ckpt-50000.index Loading commit data...
model.ckpt-50000.meta Loading commit data...
pipeline.config Loading commit data...

Training Report

date: 2020 04 27 Mon

model name: MobileNet based Single Shot MultiBox Detector with Pooling Pyramid Network

input image size: 300 x 300

entire steps: 50000

best step: 32700

batch size: 32

number of samples for training: 300

number of samples for validation: 65

metrics set: coco detection metrics

mAP@50IOU

mAP@50IOU

[32700 - best step]

mAP@50IOU: 64%

loss:

​ classification: 0.7869

​ localization: 0.4739

​ total: 1.418

Result

result1

result2

Comment

need more fire data.