About Arnaud Bastide
I am CEO, CMO and CTO at the same time. I have a project approach—listening to partners who are integrators of video surveillance solutions. I assist value-added integrators (people who resell solutions to users) with 2 tasks:
- integration between third-party products via SDKs
- project development by using computer vision, with or without ML, required on custom applications corresponding to customer needs.
A little bit about Roc4t
Roc4t—originally a consulting company in the field of video surveillance—has become an IT services and development enterprise around video surveillance based in Dubai. Roc4t works in 2 main sectors, city surveillance in smart cities, and sensitive and industrial sites around intrusion.
What were your processes before using Picsellia?
Before using Picsellia, I was creating and maintaining frameworks that served as a software base to build applications. I was seeking and testing different frameworks driven by Intel or Bosh that were open, free, scalable and maintained to develop these foundations.
How do you use Picsellia today?
At Picsellia, I use frameworks that offer pre-trained model libraries that can be used on several projects. I worked on the OpenCV challenge where my goal was to understand the processes of how to build a model. Today I'm working on a project where I have to train an Asian license plate detector.
My main objective is to achieve the best projects for my clients so that their projects can lead to another one. I am not really looking to grow the structure, I just want to satisfy my current customers. Overall, my main challenges were to find scripts, store images, collaborate with collaborators, annotate, etc.
How has Picsellia affected your productivity metrics?
Without a platform, my projects are not feasible—I don't have the time and the skills. Picsellia allowed me to reach goals that I would not have been able to without it.
It is important to highlight the simplicity of the platform, the model library is good, there are several options to train models yourself, internally via SDK or on the cloud and the OVH option looks very interesting. It's quite complete and there are several alternatives, you can go into much more detail of the models, and especially, the support is very attentive and reactive.