souhaiel1 / cnn-lstm-violence-detection

Violence Detection using a CNN-LSTM model

Hello 👋 I am Souhaiel BenSalem !

• You can reach me through my gmail or my ENS Paris Saclay academic email

• 🏫 I am currently a MSc student majoring in Mathematics, Vision and AI at ENS Paris Saclay, Paris Saclay university and a research intern at Inria Paris within the Astra-Vision group working jointly with researchers from Valeo.ai .I am also a former intern at INRIA's Chroma team, where I conducted research in (multi-agent) Reinforcement Learning.

• 📚 I obtained a MSc degree from Grenoble INP in Mobile, Autonomous Robotic Systems in 2022 , and prior to that, I earned an Engineering Degree in Robotics and Mechatronics from The university of Carthage, Tunisia, in 2021 and was ranked 1st among my peers.

• ⚡ Fun fact: I am passionate about research and I enjoy Reverse-Engineering as much as I enjoy engineering.

Connect with me:

cnn-lstm-violence-detection's Issues

Could not load fightw.hdfs

Hi,
Thanks for sharing your great work.

I'm trying to run predict_video1/2 without any success

I'm getting this warning:
2021-04-08 14:02:28.227955: W tensorflow/core/util/tensor_slice_reader.cc:95] Could not open ./fightw.hdfs: Data loss: not an sstable (bad magic number): perhaps your file is in a different file format and you need to use a different restore operator?

It may affect the results? it's not accurate at all, not even close.

Script for training

Could you please share the training script?

Tensorflow compatibility with CUDA & cuDNN

Hi,

I'm trying to run predict_video1/2.py. I got a tensorflow error of "Could not create cudnn handle: CUDNN_STATUS_INTERNAL_ERROR". It seems to be a compatibility issue. Could you tell me your version of Tensorflow, CUDA, and cuDNN?

Thank you!!!