HNHacker News
TopNewBestAskShowJobs

mendeza

325 karma · joined February 26, 2015

MLE @ DeterminedAI and Augmented Reality creator. Portfolio: http://andrewmendez.me,

PhD student at UMD College Park mendeza@umd.edu

submissionscomments
mendeza··on Create diagrams with code using Graphviz
Great tutorial! Is there a way to extract the bounding box of nodes, and the endpoints of lines? This would be useful to generate ML models to automatically parse graphs!
mendeza··on AI for generative design: Plain text to 3D Designs
Great work! I wonder if using SDF as a 3d model representation would result in higher resolution model generation (rather than voxels), some recent work in that area here: https://github.com/marian42/shapegan Adversarial Generation of Continuous Implicit Shape Representations : https://arxiv.org/abs/2002.00349
mendeza··on Ask HN: Is reading PDFs on mobile a pain?
Reading PDFs as html would Be nice, but as a ML engineer having high quality conversion would be very useful for large scale analysis and information extraction of pdf documents!
mendeza··on Becoming a high performing software developer working from your bedroom
How did you teach yourself structuring large projects and test driven development? Did you have any senior engineers mentor you when you just started, or did you just force yourself to pick important skills in software development. I ask because I work remotely a lot of the time, and I am skilled in creating a solution, but I am always worried about if it will be maintainable over time and not be overwhelming when I add features.
mendeza··on Ask HN: Who is hiring? (February 2020)
+1 for A Baltimore based AI company! I have to drive to Northern VA from Baltimore in order to do my AI work!
mendeza··on Ask HN: How do you manage your bookmarks?
I use slack to manage bookmarks. I was using Toby Chrome plugin, but after 1000 bookmarks, its search is really slow. Slack is awesome for personal use because you can easily search for links you written, and organize with new channels. Discord may be a better alternative, because there is no limit to how many messages you have.

I agree with monkeynotes that you have to be mindful on how much time you spend collecting bookmarks, I felt consumed at one point and unproductive because you skim, you save, and you find other links to the original link you are saving. Probably set a time limit to how long you spend.

mendeza··on My story as a self-taught AI researcher
I think an exciting area that can innovate the lack of data is domain randomization, and synthetic data generation.

Slides from Josh Tobin is a great introduction: http://josh-tobin.com/assets/pdf/randomization_and_the_reali...

http://josh-tobin.com/assets/pdf/BeyondDomainRandomization_T...

And a really cool project implementing synthetic generation of text in images: https://github.com/ankush-me/SynthText

mendeza··on Show HN: Screenshot Hero - iPhone app to make all photos text searchable
Reading all these comments about others having bad experiences with Tesseract makes me think there is a market for a robust, better alternative to Tesseract.
mendeza··on How to get in touch with small businesses?
You should visit indiehackers.com and see if anyone has advice on this there.
mendeza··on Show HN: Screenshot Hero - iPhone app to make all photos text searchable
I think you have to fiddle with the internal settings of the OCR package you are using to get good results. For tesseract and pytesseract, there is a whole article on improving the quality: https://github.com/tesseract-ocr/tesseract/wiki/ImproveQuali...

My guess is Apple trained on their own massive dataset and better architectures that make their systems better than off the shelf ocr.

I am working in this area and one way I see good improvement is training an object detector to first detect words in an image, then you can pass that through tesseract/OCR software. Besides that, finetuning tesseract on data you want strong performance would be the next best alternative.

Here is a cool article from dropbox how they engineering their OCR system: https://blogs.dropbox.com/tech/2017/04/creating-a-modern-ocr...

This link below is a really cool project that showcases what building your own OCR engine would look like!

https://github.com/awslabs/handwritten-text-recognition-for-...

mendeza··on Ask HN: What will become possible by 2025 in your field of expertise?
I am really excited about this possibility! I am imaging having AI augment our ability to read and conduct reading comprehension, would be amazing for kids who may be weak in certain areas to get the feedback and help they need. Thanks for sharing the link!
mendeza··on Show HN: Workouts – Learn by doing real developer projects
Would be great if there could be niche workouts of Machine Learning projects. I can see a list of machine learning projects that aspiring ML practitioners can do to build strong skills and a portfolio companies would be interested in.
mendeza··on Ask HN: What is your ML stack like?
How do you all monitor concept drift or monitoring to detect when its time to deploy a fresh model?
mendeza··on Show HN: OpenCV based multi-object-tracker with various object detection models
I actually checked out your github repo and saw your framework that enables integration of trained models into your tool for active learning. Looks awesome! https://github.com/heartexlabs/pyheartex
mendeza··on Show HN: OpenCV based multi-object-tracker with various object detection models
It would be great if your tool integrates popular object detection frameworks such as TF Object Detection API. Would save people a lot of headache installing and configuring in order to begin finetuning
mendeza··on Show HN: OpenCV based multi-object-tracker with various object detection models
This is a good idea! TF Object Detection framework takes a lot of effort to use on simple ideas, so frameworks that enable more abstraction the better!
mendeza··on Show HN: OpenCV based multi-object-tracker with various object detection models
I would check out template matching or using OpenCV feature extractors such as SIFT, SURF, ORB

Template Matching scans an image with your template, smaller image and finds the location with highest cross correlation: https://opencv-python-tutroals.readthedocs.io/en/latest/py_t...

Template matching assumes the example image wont deform or take on different orientations.

Popular Feature Extraction finds and derive more robust features that help find objects in the scenarios of distortion, orientation, and scale.

Link for feature detection and matching: https://opencv-python-tutroals.readthedocs.io/en/latest/py_t...

Pyimagesearch does a cool method where they enable template matching to work at different scales. It ignores color information (by using edge detection), which could be a critical feature if the object has specific color:

https://www.pyimagesearch.com/2015/01/26/multi-scale-templat...

mendeza··on Show HN: OpenCV based multi-object-tracker with various object detection models
Also, what is a hard but an important question to figure out is how well are these models are preforming, and will they perform well to your problem. This paper gives a lot of insight into how image size and other factors really impact performance of modern object detection: https://arxiv.org/pdf/1611.10012.pdf. How large or small your object is in your dataset is also something to consider since there is a limit to how small of an object your detector can recognize.

Also, I would recommend this repo to better understand how object detection models are evaluated as well: https://github.com/rafaelpadilla/Object-Detection-Metrics

mendeza··on Show HN: OpenCV based multi-object-tracker with various object detection models
I work a lot with object detection at my current role. Would be cool to add Detectron 2, seems that this library that has some of the state of the art(SOTA) models in object detection. https://github.com/facebookresearch/detectron2

Also the TF Object Detection API has a suite of pretrained SOTA object detection models https://github.com/tensorflow/models/tree/master/research/ob...

What makes object detection really valuable is being able to finetune a model to detect objects that are relevant to your problem. However collecting, labeling, takes a lot of time since the best way is to manually label. But 30-50 hrs of manual labeling can get you some great results when you are leveraging pretrained models. Building a tool to manage data labeling and training for finetuning object detection models would be very valuable.

mendeza··on Mathematics for Machine Learning [pdf]
I am taking CS 236 (Deep Generative Models) at Stanford and this course really helped me a lot in reviewing the math I needed for the class. This is because the chapters focus on the concepts and ideas you should know, so you don't have to waste time learning everything there is to know about linear algebra and convex optimization, but the essentials for machine learning.
mendeza··on Show HN: Instant, Self-Destructing Websites
This one is my favorite: https://sdnotes.com/facebook
mendeza··on How to do a code review
I found the DS100 course textbook and the author describes the phenomenon here: https://www.textbook.ds100.org/ch/02/design_srs_vs_big_data....

This is insightful, but how does one deal with determing if your dataset is non-random? I can imagine manually inspecting the data and using cross-validation are ways to help identify skewed datasets. Are there any other ways to test if your data is non-random?

mendeza··on Ask HN: What Neural Networks/Deep Learning Books Should I Read?
My go-to when I started in Deep Learning is Stanford's CS 231n course http://cs231n.github.io/, the lecture notes are amazing. It was instrumental when I first dove deep into Deep Learning and helped me understand all the components needed to make Convolutional Neural Networks(CNN) and Neural Networks(NN) work. I first read this and watched the lecture videos. This course was a great primer for me to understand the content and theory in Goodfellow's Deep Learning book.
mendeza··on Ask HN: How to lead a stable, happy career as a bad interviewee?
If you have trouble finding full time SWE roles after all your numerous jobs, maybe you are lacking critical skills employees want to see, or don't have personal projects that showcase technical depth.

Are you applying for junior dev roles? Maybe finding a career counselor of some sort can help. Reach out to your friends and ask for their honest advice of what they think you should do/add. Leverage your personal connections to get your first first full time job!

Ultimately, you have to persevere and be scrappy to find your first job. IMO the hardest part in your career is when you first enter the workforce. Always try to stand out and market yourself on how you would be a huge asset to the employer.

Last thought, if you are looking to join a great school/community of aspiring junior engineers and web developers, apply to Lambda School. I recommended my own brother and friends, and I have seen huge growth in their dev skills and abilities. I hope this helps!

mendeza··on Show HN: CC Search – search engine for 300M CC-licensed images
This is nice, I can see this being used for computer vision researchers and practitioners to train CNNs. Have you thought of implementing a reverse image search engine? That would be a great feature or project to work on as it can help users find images they are interested in as tags tend to be noisy.
mendeza··on A Recipe for Training Neural Networks
This recipe results in large amounts of time spent before any results occur (depending on the task you are trying to solve). Classification is an easy task to use this recipe, but when you venture into object detection or pose estimation, data collection, labeling, and setting up training and evaluation infrastructure is much more complex.
mendeza··on Building a fashion search engine with deep learning
Do you think if cameras can capture real depth values, or capture volumetric information, could you at least prevent shoppers buying clothes that will not likely fit. I don’t think personalizes fit would be solved, but if you could measure dimensions and include that in recommendation engines, it could provide more insight.
mendeza··on Show HN: MIT DeepTraffic 75.01 MPS Solution
The .js file is not readable, I am interested in seeing the architecture that lead to the solution. It seems that much of the file is model weights.
mendeza··on Ask HN: Can a machine learning engineer switch to a web development career?
My question is why would you want to switch to web development. IMO ML Engineering is an exciting field, I am curious to hear what your struggles are or qualms with the field.
mendeza··on Pytext: A natural language modeling framework based on PyTorch
I love textacy, it has soo much out of the box. Topic modeling, topic extraction, summarization, and its built on top of Spacy. https://github.com/chartbeat-labs/textacy
← PreviousPage 2 of 7Next →