.H5
googlenet_weights.h5 (47.58 MB)
.H5
EdgeRender_window10batch25LR0.001beta600LSTM256Dropout0.50.5.h5 (56.54 MB)
.H5
GradmagSynCar_window10batch25LR0.001beta600LSTM256Dropout0.250.25.h5 (56.54 MB)
.H5
SynCar_window10batch25LR0.001beta600LSTM256Dropout0.250.25.h5 (56.54 MB)
.H5
SynPhoReal_window10batch25LR0.001beta600LSTM256Dropout0.50.5.h5 (56.54 MB)
.H5
SynPhoRealTex_window15batch15LR0.001beta600LSTM256Dropout0.50.5.h5 (56.54 MB)
1/0
GoogleNet weights trained on the Places dataset for Keras.
Version 2 2020-02-17, 05:36
Version 1 2019-11-23, 07:23
dataset
posted on 2020-02-17, 05:36 authored by Debaditya Acharya, KOUROSH KHOSHELHAMKOUROSH KHOSHELHAM, STEPHAN WINTERSTEPHAN WINTERThis file (googlenet_weights.h5) contains the initial weights of GoogleNet (aka v1) trained on the Places dataset, and is used for fine-tuning task related to pose regression networks. The file was downloaded from https://github.com/kentsommer/keras-posenet. Adding to the repository is only for the purpose of eliminating dependency on external URLs.
Other files contain the weights of the final trained model from our experiments of the paper Recurrent BIM-PoseNet:
SynCar - Weights of model fine-tuned on Synthetic Cartoonish images.
SynPhoReal - Weights of model fine-tuned on Synthetic photo-realistic images.
SynPhoRealTex - Weights of model fine-tuned on Synthetic photo-realistic textured images.
GradmagSynCar - Weights of model fine-tuned on synthetic gradmag of SynCar images.
EdgeRender - Weights of model fine-tuned on Synthetic edge render images.
Other details in the name of the weight files describes the parameters, such as window length, learning rate, batch, ...., etc.
This is a support file for the code available at https://github.com/debaditya-unimelb/RecurrentBIM-PoseNet.
