Commits

Branch selector

 master

User selector

All users

Datepicker

All time

Commit History

Commits on Jun 27, 2019

• script for pb_to_uff conversion. Also a c++ from nvidia added to convert uff to .plan (aka .engine)

  mpkuse

  committed
  a50abe1

  Copy full SHA for a50abe1

Commits on Jun 18, 2019

• added a new utility to convert .h5 to tensorflow protobuf (frozen graph). readme updated to reflect this

  mpkuse

  committed
  21c980f

  Copy full SHA for 21c980f

• prediction from frozen (.pb)

  mpkuse

  committed
  c051c48

  Copy full SHA for c051c48

Commits on Jun 4, 2019

• added various backend cnn for references

  mpkuse

  committed
  1e428d6

  Copy full SHA for 1e428d6

• mobilenetv2, new utilities for show modelinfo, edit model inputs, and convert model to uff

  mpkuse

  committed
  dd575d2

  Copy full SHA for dd575d2

Commits on May 31, 2019

• tensorrt on desktop pc and test script that works on desktop for uffparser

  mpkuse

  committed
  b4125b2

  Copy full SHA for b4125b2

Commits on May 30, 2019

• Got TensorRT working.

  Show description for 0d8bbfc
  mpkuse

  committed
  0d8bbfc

  Copy full SHA for 0d8bbfc

Commits on May 29, 2019

• updated readme and make the utility to convert keras h5 modelarch_and_weight files to .uff

  mpkuse

  committed
  c23f23b

  Copy full SHA for c23f23b

• normalization example

  mpkuse

  committed
  d125d56

  Copy full SHA for d125d56

• moved to HDF5 model arch and weight storing, eventually move away from json

  mpkuse

  committed
  8feff13

  Copy full SHA for 8feff13

Commits on May 27, 2019

• make the script test_kerasmodel_to_pb.py compatible with my docker image mpkuse/kusevisionkit:tfgpu-1.12-tensorrt-5.1

  mpkuse

  committed
  d62c862

  Copy full SHA for d62c862

• unit test on how to convert keras model to tensorflow proto-binary (pb). test_kerasmodel_to_pb.py

  mpkuse

  committed
  8ee2795

  Copy full SHA for 8ee2795

Commits on May 1, 2019

• GhostVLAD implementation.

  Show description for 51e9302
  mpkuse

  committed
  51e9302

  Copy full SHA for 51e9302

Commits on Apr 25, 2019

• some prediction utils

  mpkuse

  committed
  b87adc7

  Copy full SHA for b87adc7

Commits on Apr 23, 2019

• SIGINT handler; some more configs in one place easy to tune

  mpkuse

  committed
  f0b1bc8

  Copy full SHA for f0b1bc8

Commits on Apr 22, 2019

• the bug on keras getting stuct on 1st epoch fixed

  mpkuse

  committed
  079992f

  Copy full SHA for 079992f

Commits on Apr 17, 2019

• a easy to use sequence generator in keras

  mpkuse

  committed
  404ec4c

  Copy full SHA for 404ec4c

• bug fixes in rendering with pitssburg

  mpkuse

  committed
  6b6a28c

  Copy full SHA for 6b6a28c

Commits on Dec 7, 2018

• added scripts to convert matlab model to keras and run original netvlad

  mpkuse

  committed
  6338b59

  Copy full SHA for 6338b59

Commits on Dec 6, 2018

• cleanup

  mpkuse

  committed
  564a5fd

  Copy full SHA for 564a5fd

Commits on Nov 30, 2018

• GFLOPs, memory, nparams computations of a network

  mpkuse

  committed
  e96e608

  Copy full SHA for e96e608

Commits on Nov 6, 2018

• model.png for various pretrained models in unit_tests. also added positve set deviation loss and metric

  mpkuse

  committed
  ce9f687

  Copy full SHA for ce9f687

• moved all my custom loss function and eval metrics into CustomLosses.py

  mpkuse

  committed
  92eea99

  Copy full SHA for 92eea99

Commits on Oct 25, 2018

• trying attention

  mpkuse

  committed
  867174b

  Copy full SHA for 867174b

Commits on Oct 17, 2018

• acceptable data augmentation and some changes to triplet loss

  mpkuse

  committed
  81793c0

  Copy full SHA for 81793c0

Commits on Oct 16, 2018

• triplet loss and verified pairwise loss. turns out there was a bug in loss :(

  mpkuse

  committed
  0b59f94

  Copy full SHA for 0b59f94

Commits on Oct 15, 2018

• noveou_train_netvlad_v3.py using mobilenet and deeper features. no upsampling. uses generator and fully tested loss function and validation metric

  mpkuse

  committed
  a105ea8

  Copy full SHA for a105ea8

Commits on Oct 11, 2018

• training netvlad with my loss and 2 up-sampling convolution layers each has its batchnorm

  mpkuse

  committed
  09b9914

  Copy full SHA for 09b9914

Commits on Oct 10, 2018

• data augmentation with external library. got rid of custom dataaugmentation. added dependence on imgaug

  mpkuse

  committed
  cdd03dc

  Copy full SHA for cdd03dc

Commits on Oct 8, 2018

• v2 is working as expected.

  mpkuse

  committed
  9e15a8d

  Copy full SHA for 9e15a8d

Commits on Oct 7, 2018

• v2 of noveou contain timedistributed model which i think is more modular and consumes less memory

  mpkuse

  committed
  2d0af01

  Copy full SHA for 2d0af01

Commits on Oct 5, 2018

• Tensorflow1.11 and Keras

  Show description for 3600af6
  mpkuse

  committed
  3600af6

  Copy full SHA for 3600af6

Commits on Aug 20, 2018

• added a class interface

  mpkuse

  committed
  c23be6b

  Copy full SHA for c23be6b

• minor edit in standalone code

  mpkuse

  committed
  004adf0

  Copy full SHA for 004adf0

• added a standalone test code. Easy to use model in your own application

  mpkuse

  committed
  eac7c76

  Copy full SHA for eac7c76

Pagination

Previous

Next