Machine Box Classificationbox
AI
image classification
SHOW HN2018
Machine Box Classificationbox
Quickly build a production ready image classifier from your images
blog.machinebox.io#image classifier#machine learning#docker#image recognition#training data
How HN took it
1 positive, 5 neutral or asking, 0 mixed, 3 critical, out of 9 comments about the project. Thread score -0.09.
Performance: 0 for, 2 againstExecution: 1 for, 0 againstIdea: 0 for, 1 against
The thread, comment by comment
Each row is a top-level comment and its replies, in HN's order. Select a cell to open that comment on HN.
positive neutral or mixed critical about another commenter or something else the maker
What HN said
What people liked
Looks great and easy to try
From 3 comments
This looks great! But not entirely clear when looking at the site, what would be the pricing to classify say 20,000 images from 20 different folders?
pitchups on HNLooks great! Uploaded to https://notebooks.azure.com/smortaz/libraries/hvasslab-TF-tutorials ready to clone & run...
smortaz on HNThis is neat, but sadly useless for what I'd imagine many of us have. Pictures of people we want to classify. Possibly of events. Classifying events is often most easily done by…
taeric on HN
Objections
Accuracy not production-grade; classifiers confuse similar images
From 2 comments
It is interesting that the words "production ready" can be used in connection to image classifiers. Google is unarguably the leader of this in the world, and yet Google photos…
dmitrygr on HNAs I said, there are some pictures where the only thing in them are my kids. Lighting was such that they are literally all that is in the picture. And I'm talking about 3 month…
taeric on HN
Base64 encoding adds about a third to image size
From 2 comments
No, base64 is normally ~30% larger than binary to represent the same data.
tonyhb on HNBase64 doesn't compress at all, in fact it's only 75% as efficient as binary, meaning that encoding a file into Base64 increases its size by a third. Base64 uses a 64 character…
JimDabell on HN
Unclear whether data leaves the user's system during training
From 1 comment
I think it would be worthwhile for HN to come up with some sort of rule or strong guideline at least that says if you are going to post a Show HN item, you should pop in and…
natch on HN
Requests
Show pricing for large image sets
From 1 comment
This looks great! But not entirely clear when looking at the site, what would be the pricing to classify say 20,000 images from 20 different folders?
pitchups on HN
Tutorial covering custom categories not in standard models
From 1 comment
I think it would be worthwhile for HN to come up with some sort of rule or strong guideline at least that says if you are going to post a Show HN item, you should pop in and…
natch on HN
GPU support in the Docker image
From 1 comment
Is that Docker image capable of utilizing a GPU?
amelius on HN
Questions
How many images are needed to train well
From 1 comment
Do I need a lot of images for this to work? One of the biggest barriers to entry in machine learning is not having enough sample data to train on. I don’t have 10,000 images of…
statictype on HN
Why images are sent as base64
From 1 comment
>Teaching essentially involves opening each image, converting it into a base64 string, and submitting it to the /classificationbox/teach API endpoint in a request that resembles…
aisofteng on HN
Alternatives named
Google Photos image search
From 1 comment
It is interesting that the words "production ready" can be used in connection to image classifiers. Google is unarguably the leader of this in the world, and yet Google photos…
dmitrygr on HN
Google and Amazon image recognition services
From 1 comment
As I said, there are some pictures where the only thing in them are my kids. Lighting was such that they are literally all that is in the picture. And I'm talking about 3 month…
taeric on HN