Bought one for my kiddo.
He’s just a baby but still needs a unique email for me to create him accounts (eg frequent flyer, global entry) and I had previously been using a +variant on my gmail.
Having a vanity email alias seems more fun.
3,589 karma · joined April 5, 2012
Past: founder of Hatchlings, www.hatchlings.com
Contact: https://twitter.com/braddwyer (DMs Open)
He’s just a baby but still needs a unique email for me to create him accounts (eg frequent flyer, global entry) and I had previously been using a +variant on my gmail.
Having a vanity email alias seems more fun.
Roboflow is the fastest way to use computer vision in production. We help developers give their software the sense of sight. Our end-to-end platform[1] provides tooling for image collection, annotation, dataset exploration and curation, training, and deployment.
Over 100k engineers (including engineers from 2/3 Fortune 100 companies) build with Roboflow. And we now host the largest collection[2] of open source computer vision datasets and pre-trained models[3].
We have several openings available, but are primarily looking for strong technical generalists who want to help us democratize computer vision and like to wear many hats and have an outsized impact. (We especially love hiring past and future founders.)
We're hiring 3 full-stack engineers this quarter, a back-end focused engineer to work primarily on our APIs and SDKs, and we're also looking for an infrastructure engineer.
[1]: https://docs.roboflow.com/?ref=whoishiring1222
[2]: https://blog.roboflow.com/computer-vision-datasets-and-apis/...
Re synthetic data specifically, we've written a couple of how-to guides for creating data from context augmentation[5], Unity Perception[6], and Stable Diffusion[7] & are talking to some others as well; it seems like a natural integration point (and someplace where we don't need to reinvent the wheel).
[1] https://docs.roboflow.com/rest-api
[2] https://github.com/SkalskiP/make-sense/pull/298
[3] https://github.com/ultralytics/yolov5/discussions/10425
[4] https://docs.roboflow.com/train/pro-third-party-training-int...
[5] https://blog.roboflow.com/how-to-create-a-synthetic-dataset-...
[6] https://blog.roboflow.com/unity-perception-synthetic-dataset...
[7] https://blog.roboflow.com/synthetic-data-with-stable-diffusi...
We’d like to make this as easy as possible for folks to use while giving proper credit to the users who did the hard work labeling and sharing these datasets.
~~Could you file an issue on the repo for the download issue you encountered? We just pushed an update to our python package earlier today that might be related.~~
~~Update: I went ahead and filed this issue[0] so the relevant engineer will see it when they wake up.~~
Update 2: I pushed a fix to our backend API & verified it's working by running that download script in a Colab notebook.
> What are the exact attribution instructions?
Creative Commons has some guidance[1] on how they recommend citing attribution.
[0] https://github.com/roboflow/roboflow-100-benchmark/issues/31
[1] https://wiki.creativecommons.org/wiki/Recommended_practices_...
We originally considered “fixing” the labels on these datasets by hand, but ultimately decided that label error is one of the challenges “real world” datasets have that models should work to become more robust against. There is some selection bias in that we did make sure that the datasets we chose passed the eye test (in other words, it looked like the user spent a considerable amount of time annotating & a sample of the images looked like they labeled some object of interest).
For aerial images in particular my guess would be that these models suffer from the “small object problem”[1] where the subjects are tiny compared to the size of the image. Trying a sliding window based approach like SAHI[2] on them would probably produce much better results (at the expense of much lower inference speed).
* Blog Post: https://blog.roboflow.com/roboflow-100/
* Paper: https://arxiv.org/abs/2211.13523
* Github: https://github.com/roboflow-ai/roboflow-100-benchmark
At Roboflow, we've seen users fine-tune hundreds of thousands of computer vision models on custom datasets.
We observed that there's a huge disconnect between the types of tasks people are actually trying to perform in the wild and the types of datasets researchers are benchmarking their models on.
Datasets like MS COCO (with hundreds of thousands of images of common objects) are often used in research to compare models' performance, but then those models are used to find galaxies, look at microscope images, or detect manufacturing defects in the wild (often trained on small datasets containing only a few hundred examples). This leads to big discrepancies in models' stated and real-world performance.
We set out to tackle this problem by creating a new set of datasets that mirror many of the same types of challenges that models will face in the real world. We compiled 100 datasets from our community spanning a wide range of domains, subjects, and sizes.
We've benchmarked a couple of models (YOLOv5, YOLOv7, and GLIP) to start, but could use your help measuring the performance of others on this benchmark (check the GitHub for starter scripts showing how to pull the dataset, fine-tune models, and evaluate). We're very interested to learn which models do best in which real-world scenarios & to give researchers a new tool to make their models more useful for solving real-world problems.
Roboflow is the fastest way to use computer vision in production. We help developers give their software the sense of sight. Our end-to-end platform[1] provides tooling for image collection, annotation, dataset exploration and curation, training, and deployment.
Over 100k engineers (including engineers from 2/3 Fortune 100 companies) build with Roboflow. And we now host the largest collection[2] of open source computer vision datasets and pre-trained models[3].
We have several openings available, but are primarily looking for strong technical generalists who want to help us democratize computer vision and like to wear many hats and have an outsized impact. (We especially love hiring past and future founders.)
We're hiring 3 full-stack engineers this quarter, a back-end focused engineer to work primarily on our APIs and SDKs, and we're also looking for an infrastructure engineer with Elasticsearch experience.
[1]: https://docs.roboflow.com/?ref=whoishiring1222
[2]: https://blog.roboflow.com/computer-vision-datasets-and-apis/...
* Paper: https://arxiv.org/abs/2211.13523
* Github: https://github.com/roboflow-ai/roboflow-100-benchmark
At Roboflow, we've seen users fine-tune hundreds of thousands of computer vision models on custom datasets.
We observed that there's a huge disconnect between the types of tasks people are actually trying to perform in the wild and the types of datasets researchers are benchmarking their models on.
Datasets like MS COCO (with hundreds of thousands of images of common objects) are often used in research to compare models' performance, but then those models are used to find galaxies, look at microscope images, or detect manufacturing defects in the wild (often trained on small datasets containing only a few hundred examples). This leads to big discrepancies in models' stated and real-world performance.
We set out to tackle this problem by creating a new set of datasets that mirror many of the same types of challenges that models will face in the real world. We compiled 100 datasets from our community spanning a wide range of domains, subjects, and sizes.
We've benchmarked a couple of models (YOLOv5, YOLOv7, and GLIP) to start, but could use your help measuring the performance of others on this benchmark (check the GitHub for starter scripts showing how to pull the dataset, fine-tune models, and evaluate). We're very interested to learn which models do best in which real-world scenarios & to give researchers a new tool to make their models more useful for solving real-world problems.
We used something similar to build a “similar articles” feature & it gave us de-duplication essentially for free.
Roboflow is the fastest way to use computer vision in production. We help developers give their software the sense of sight. Our end-to-end platform[1] provides tooling for image collection, annotation, dataset exploration and curation, training, and deployment.
Over 100k engineers (including engineers from 2/3 Fortune 100 companies) build with Roboflow. And we now host the largest collection[2] of open source computer vision datasets and pre-trained models[3].
We have several openings available, but are primarily looking for strong technical generalists who want to help us democratize computer vision and like to wear many hats and have an outsized impact. (We especially love hiring past and future founders.)
We're hiring 3 full-stack engineers this quarter and we're also looking for an infrastructure engineer with Elasticsearch experience.
[1]: https://docs.roboflow.com
[2]: https://blog.roboflow.com/computer-vision-datasets-and-apis/
If you've seen a paper that verifies this I'd love to see it. The early layers of the network detecting simple features like lines, curves, textures, shapes, colors, etc could still benefit from what they learn on ImageNet even if the features later in the network are not similar.
FWIW I have not yet seen a model starting from a pre-trained COCO checkpoint that does worse than random initialization.
The intuition for this is that lines, colors, textures, shapes are all general concepts that can be learned from a different domain & used in the earlier layers of the model to build up to more complex features.
To self-host just run
npx @roboflow/inference-server
Then you can POST an image at any of the models to localhost:9001 eg
base64 yourImage.jpg | curl -d @- "http://localhost:9001/some-model/1?api_key=xxxx"
And you get back JSON predictions.
There are also client libs[2] and sample code[3] for pretty much any language you might want to use it in. You can also run any of the models directly in a browser with WebGL[4], in a native mobile app[5], or on an edge device[6].
[1] https://universe.roboflow.com
[2] https://github.com/roboflow-ai/roboflow-python
[3] https://github.com/roboflow-ai/roboflow-api-snippets
[4] https://docs.roboflow.com/inference/web-browser
[5] https://docs.roboflow.com/inference/mobile-ios-on-device
My opinion is that if the outcome is the same (a view from above is made publicly available) the how isn't a "large difference" from a moral perspective (which is what you were asking about).
Personally, I think it's fine, and don't have a problem with people flying over my backyard & collecting images just like I don't have a problem with satellites adding images of my yard to Google Earth. I think it's pretty cool that I can zoom in and see my car in my driveway as seen by a satellite flying around the Earth at 17,000 mph.
I'm not really an expert so I just labeled the things I thought looked like solar panels. It's mostly meant to be an example -- if you want to use something like this in a production environment you should definitely spend more than the ~2 hours on image collection, dataset labeling, and training that I did for this!
[1] https://universe.roboflow.com/brad-dwyer/aerial-solar-panels...
Whether this is or is not "right" in the age of ubiquitous drone availability is up to us (via our government) to decide/change.
Disclaimer: IANAL and before someone starts a company based on this, they should definitely consult one.
I would be thrilled to have someone pick this repo up and run with it. We're really focused on the developer tools and core infrastructure behind the computer vision part. We don't have the bandwidth to capitalize on all these opportunities (and this is only in the aerial space which is a tiny portion of the use-cases we've already seen for the technology; engineers at over 2/3 of the Fortune 100 have used Roboflow for things in every industry from retail to broadcasting to manufacturing to agriculture to healthcare).
Having meaningful businesses built on top of our tools is a great sign that we're building something useful (and we'd be ecstatic to help them succeed if they want to become a customer)!
Related: we recently released[2] semantic search for custom datasets. You can use it to find all sorts of weird stuff in benchmark datasets like MS COCO[3] used to train many computer vision models.
If folks are interested I can write up a "how we made it" post describing the behind the scenes.
Roboflow is the fastest way to use computer vision in production. We help developers give their software the sense of sight. Our end-to-end platform[1] provides tooling for image collection, annotation, dataset exploration and curation, training, and deployment.
Over 100k engineers (including engineers from 2/3 Fortune 100 companies) build with Roboflow. And we now host the largest collection[2] of open source computer vision datasets and pre-trained models[3].
We have several openings available, but are generally looking for strong technical generalists who want to help us democratize computer vision and like to wear many hats and have an outsized impact. (We especially love hiring past and future founders.)
[1]: https://docs.roboflow.com
[2]: https://blog.roboflow.com/computer-vision-datasets-and-apis/