I have been trying out a TensorFlow application called DeepLab that uses deep convolutional neural nets (DCNNs) along with some other techniques to segment images into meaningful objects and than label what they are. Using a script included in the DeepLab GitHub repo, the Pascal VOC 2012 dataset is used to train and evaluate the model. One of the results is shown above. It has managed to extract some pretty ugly furniture from a noisy background quite nicely. Here are couple more examples:
The software has done a nice job of extracting the foreground objects in another very noisy scene.
The person in the background is picked up pretty nicely here – I didn’t even notice the person at first.
Incidentally, to get the local_test.sh to work on Ubuntu 16.04 I had to change the call to download_and_convert_voc2012.sh to use bash instead of sh otherwise it generated an error. Also, I needed to install cuDNN 7.0.4 for Cuda 9.0 rather than cuDNN 7.1.1 in order to get the Jupyter notebook example operating.
What I would like to do now is to create an rt-ai Edge Stream Processing Element (SPE) based on this code to act as a preprocessor stage in order to isolate and identify salient objects in a video stream in real time. One of my interests is understanding behaviors from video and this could be a valuable component in that pipeline by allowing later stages to focus on what’s important in each frame.
Any time I start a project I always wonder if I am just reinventing the wheel. After all, there is so much software out there (on GitHub and others) that almost everything already exists in some form. The most obvious analog to rt-ai Edge is Apache NiFi and Apache MiNiFi. NiFi provides a very rich environment of processor blocks and great tools for joining them together to create stream processing pipelines. However, there are some characteristics of NiFi that I don’t particularly like. One is the reliance on the JVM and the consequent garbage collection issues that mess up latency guarantees. Tuning a NiFi installation can be a bit tricky – check here for example. However, many of these things are the price that is inevitably paid for having such a rich environment.
rt-ai Edge was designed to be a much simpler and lower overhead way of creating flexible stream processing pipelines in edge processors with low latency connections and no garbage collection issues. That isn’t to say that an rt-ai Edge pipeline module could not be written using a managed memory language if desired (it certainly could) but instead that the infrastructure does not suffer from this problem.
In fact, rt-ai Edge and NiFi can play together extremely well. rt-ai Edge is ideal at the edge, NiFi is ideal at the core. While MiNiFi is the NiFi solution for embedded and edge processors, rt-ai Edge can either replace or work with MiNiFi to feed into a NiFi core. So maybe it’s not a case of reinventing the wheel so much as making the wheel more effective.
The “rt” part of rt-ai doesn’t just stand for “richardstech” for a change, it also stands for “real-time”. Real-time inference at the edge will allow decision making in the local loop with low latency and no dependence on the cloud. rt-ai includes a flexible and intuitive infrastructure for joining together stream processing pipelines in distributed, restricted processing power environments. It is very easy for anyone to add new pipeline elements that fully integrate with rt-ai pipelines. This leverages some of the concepts originally prototyped in rtndf while other parts of the rt-ai infrastructure have been in 24/7 use for several years, proving their intrinsic reliability.
Edge processing and control is essential if there is to be scalable use of intelligent IoT. I believe that dumb IoT, where everything has to be sent to a cloud service for processing, is a broken and unscalable model. The bandwidth requirements alone of sending all the data back to a central point will rapidly become unworkable. Latency guarantees are difficult to impossible in this model. Two advantages of rt-ai (keeping raw data at the edge where it belongs and only upstreaming salient information to the cloud along with minimizing required CPU cycles in power constrained environments) are the keys to scalable intelligent IoT.
rt-ai Edge is a new concept in edge processing that makes it easy for anyone to build AI and ML enhanced stream processing pipelines in order to close the local loop and offload communications networks and the cloud. Semantic extraction of meaningful data from raw data feeds at the edge ensures that the core only has to deal with actionable information, not noise. rt-ai Edge leverages hardware acceleration within embedded devices to filter raw data into highly salient messages for higher level processing.
rt-ai Edge is in active development right now.