Please enable JavaScript.
Coggle requires JavaScript to display documents.
Perception (Data (Imagenet (object classification, detection, Used for pre…
Perception
Data
Imagenet
object classification, detection
-
150,000 images, 1000 categories
-
KITTI
Many different kinds, stereo, object detection, Lidar, ...
detection / 10,000 images
PASCAL_VOC
cassification / 17,125 images / 20 categories
-
-
CITISCAPES
urban scenes, segmentation, 20,000 coarse, 5,000 fine images
-
Embedding
-
-
C-code conversion? (Since Tensorflow ckpt files include all meta data, so I guess it would be not that complex. Independent to development environment.)
-
-
Project
-
Network
-
-
-
Can use an existing network with well-pre-trained weights, and touch later layers for one's purpose.
If we totally have totally new architecture, we need to train using the Imagenet.
folder structure
root
-
-
-
-
-
-
tools # train, test start