PM: Snapchat's face detection is really good, how can we improve ours?
Me: Well they spend $150M buying a company for that face detection tech...
PM: How can we improve ours?
Me: Buy snapchat?

So I'm working on a computer vision project that grabs video from my webcam and detects faces in each frame. Earlier yesterday morning I was capping out at 30 frames per second, which is what I believe to be the max for my webcam. As the day became night and I was wrapping up my work on a portion of the project, I noticed that my newly compiled version was only getting around 8 frames per second. Confused, I looked into my frame grabbing + face detection code.

"Maybe I can only detect faces in a certain region of the image, based on where the face was in the previous image?" No, still 8fps. Hmm.

"How about I lower the resolution of the image, that would definitely help!" I tried that, but no speed boost came either. What??

I start to dig deeper. Maybe I'm not linking my libraries correctly, and it's using an older library I compiled. So I recompile that. Nothing.

"Am I low on resources?" I close out of all my other apps. Nothing.
Okay, wtf. Now I just comment out the face detection code entirely, and only grab webcam frames.

8fps. ?????

Suddenly, I get an idea. I get out of my desk, walk over to the doorway of my room, and flip the light on. I sit back down, and run my code.

30fps.

The stupid webcam switches to "night mode" when it detects low light, which restricts its ability to output frames at high speed and caps at 8fps. Damn, I felt like a fool 😂

"Oh c'mon, my 14 yr old son can change background color of entire website and you're charging me extra for simple face detection feature?"

When you realize your face detection doesn't work because you used the cat face cascade

python machine learning tutorials:
- import preprocessed dataset in perfect format specially crafted to match the model instead of reading from file like an actual real life would work
- use images data for recurrent neural network and see no problem
- use Conv1D for 2d input data like images
- use two letter variable names that only tutorial creator knows what they mean.
- do 10 data transformation in 1 line with no explanation of what is going on
- just enter these magic words
- okey guys thanks for watching make sure to hit that subscribe button

ehh, the machine learning ecosystem is burning pile of shit let me give you some examples:

- thanks to years of object oriented programming research and most wonderful abstractions we have "loss.backward()" which have no apparent connection to model but it affects the model, good to know
- cannot install the python packages because python must be >= 3.9 and at the same time < 3.9
- runtime error with bullshit cryptic message
- python having no data types but pytorch forces you to specify float32
- lets throw away the module name of a function with these simple tricks:
"import torch.nn.functional as F"
"import torch_geometric.transforms as T"
- tensor.detach().cpu().numpy() ???
- class NeuralNetwork(torch.nn.Module):
def __init__(self):
super(NeuralNetwork, self).__init__() ????
- lets call a function that switches on the tracking of math operations on tensors "model.train()" instead of something more indicative of the function actual effect like "model.set_mode_to_train()"
- what the fuck is ".iloc" ?
- solving environment -/- brings back memories when you could make a breakfast while the computer was turning on
- hey lets choose the slowest, most sloppy and inconsistent language ever created for high performance computing task called "data sCieNcE". but.. but. you can use numpy! I DONT GIVE A SHIT about numpy why don't you motherfuckers create a language that is inherently performant instead of calling some convoluted c++ library that requires 10s of dependencies? Why don't you create a package management system that works without me having to try random bullshit for 3 hours???
- lets set as industry standard a jupyter notebook which is not git compatible and have either 2 second latency of tab completion, no tab completion, no documentation on hover or useless documentation on hover, no way to easily redo the changes, no autosave, no error highlighting and possibility to use variable defined in a cell below in the cell above it
- lets use inconsistent variable names like "read_csv" and "isfile"
- lets pass a boolean variable as a string "true"
- lets contribute to tech enabled authoritarianism and create a face recognition and object detection models that china uses to destroy uyghur minority
- lets create a license plate computer vision system that will help government surveillance everyone, guys what a great idea

I don't want to deal with this bullshit language, bullshit ecosystem and bullshit unethical tech anymore.

So I've finally implemented face detection in my discord bot.

@Wisecrack regarding the openCV training, I'm still taking positive and negative samples, so it'll be another while.

Also, why is post @ing not a thing

(Part 1/2?)

Ohhh my god am I furious and this one's a gem.

Also I'm gonna namespoil all of the entities in my post. If this is against rant rules I'll reframe it.

So the story starts over an year ago. Me, being in a bad place, where I couldn't do a job due to external issues, wanted to try out an internship. Thought I could pull off a 5 hour shift and then attend to my problems.

THE INTERNSHALA ARC:

I apply to a bunch of applications on Angel, Internshala and Indeed.

I was contacted by a few handful of these places. One of them was called "ARCHITECTA SOFTWARE SOLUTIONS". These guys had arranged an online aptitude test for me which I promptly took.

I looked up this company and they seemed like a pretty okay big firm from the outset but didn't have many reviews on Glassdoor and likes of such. (first red flag). Post aptitude test, I was quite sure I fucked up and wouldn't get further contact. Surprisingly, a person from the company sends me his Whatsapp number over chat and asks me to save it. The message is worded like a bulk email (Starting with Hello everyone!!) which I thought was quite odd since the interaction from these platforms has always been a person-to-person contact for me. Since Internshala showed that only around 40 people applied for the position I was quite intrigued but attributed this to my lack of exp in internship operations.

THE WHATSAPP ARC:

I was contacted by the number on WhatsApp saying that they'd be interested in moving forward and I gave them my work experience details.

The person sends me over a development assignment to complete within a few days. The assignment consists of massive scope of details. I'm talking production level concept and implementation. Asks to me implement a custom emotion detection CV model (worded as "emotion camera" lmao), generate a 3d model (specified nowhere and expects to implement a mono-ocular system for the curious) and deploy it over AWS with a website to go along with it and also host that. The website should contain a VR ("360 rotatable") view that can explore the depth-map ("not worded as depth-map") of the face. My first assumption was that they had picked this work up for outsourcing and didn't bother to chip off parts so as to create an assignment out of it (I know very optimistic).

So I shoot it at him on WhatsApp asking which parts of the assignment should I do?

Him: So, which parts CAN you do?

I thought of it as an HR thing.

Me: I could do most of it but given the time-frame of the assignment and my applied position as a web developer it is perhaps out of scope for my application.

Him: Don't worry about the assignment. You can submit when you complete the whole assignment.

I was visibly angry over the stupidity of this man.

Me: This task is a Full-Stack + CV + VR task. It will take over two months to get working. Am I supposed to work on it for that long for an assignment?

Him: Okay just do the basic functionalities like add to cart. But also try to do the camera thing before next week.

At this point I'm sure that they are having trouble handling an eager client and they're offloading work to interns. So I do only the backend and minimal frontend and submit the assignment (a 2 day job done over a weekend).

Nothing. Empty. No messages since then. I tried sending in a Whatsapp message on the application and how to proceed. Then, if I could get to know if I have been rejected. Nothing.

And all this time I can clearly see the account is active as it pushes pretentious motivational quotes over it's Whatsapp status.

Hello mostly human community.
Are there any recommended Haarcascade models for OpenCV for face detection apart from the ones shipped with it?

So apple event today was 2 hours long to say....

MOAR screen for MOAR dollar?! 💵💵😂

The only feature I want is the fall detection for when I face-palm now....

!rant
My dad wants to add security cameras in our home. I kind of want to go ahead and add cool stuff(like automation or something) to our house using this opportunity.
What I am thinking of right now is salvaging an old desktop with an Nvidia GPU to add face recognition and motion detection to the mix. I am also thinking I could get an Alexa echo dot or something similar and hook everything up. Another idea is using owncloud to create my own cloud.
What do you guys think? Any ideas or suggestions? Maybe a cheaper way to do stuff?

Any suggestions for tutorials / tips on doing facial identification? I want to identify a face with a label. Most of the shit I can find is face detection which is not what I’m looking for