Small Grants: To name an asteroid after Nouns


We are a group of astronomy and AI enthusiasts who love Nouns. In the field of astronomy, the highest glory is to officially name an asteroid after our own name - so why not name an asteroid after “Nouns” and make it Nounish?

Our goal

To name an asteroid, we have to be the first one to spot it and submit the finding to the International Astronomical Union (IAU). After the asteroid is confirmed, people who found it at first will gain the right to name it. Remember: once the name of the specific asteroid is decided, it will be permanent. The Nounish asteroid will exists forever, showing the great impact of our community and leaving our footsteps in the deep universe. It will be something that can last for hundreds and thousands of years.


In the past, people used their eyes to detect asteroids. Recently, some large observatories started to use computer programs to go through the huge datasets generated by the telescopes. However, these automated programs are not open source, and most astronomy amateurs are still using their eyes to do the task.

We believe that using the latest technology of deep learning and computer vision will boost the overall efficiency of finding asteroids, and we believe that the tools to do such tasks should be public available. Thus, we plan to put our experience in astronomy and deep learning together and do something big for the Nouns community - be the first to find an asteroid and name it after Nouns.

How we will benefit Nouns

Firstly, as stated above, it is highly possible for the community to have our own asteroid. We cannot guarantee it, yet we will do our best to reach the goal. We promise that the model we use will be published open source, and will be noted that the work is supported by Nouns. What’s more, we believe it will be a great promotion opportunity, both for the academic field and for the public.

Our plan

We planned to do the project for 9 months. The confirmation process of IAU usually lasts for 3 years, so the final output will probably come after the process.

  • Month 1~3: fully research the area and build up the training datasets
  • Month 4~6: train the model and start to use the model to detect asteroids
  • Month 7~9: improve the performance of the model and try to make the final detection
  • If everything goes well, we need to communicate with IAU for three years in order to get the final name.


Though the goal is ambitious, it won’t cost too much. We will ask for 8 ETH from Nouns to initiate and finish the program:

  • 3 ETH to build up a team of professionals and students to work full-time or part-time for the project
  • 1 ETH for the cost to build up the training datasets
  • 3 ETH to rent a high performance server to train the model (with good GPU, >16 cores of CPU, and >32 GBs of memory), for at least half a year
  • 0.5 ETH for team building
  • 0.5 ETH for the three-year-communication with IAU, to name the asteroid

If we really name an asteroid after Nouns and it is confirmed by IAU, we would like to ask for 2 ETH retro fund. It will definitely be a worthy investment.



these are the kind of grants i used to dream about seeing in Nouns! Lets name an asteroid


Thanks for your support! @davinoyesigye

For those tho like my idea:

Email Address:

Wallet Address: 0x49ba6B8A42303FE1E91D25460f769d200e5f6f85

1 Like

Super excited to see this! Even though I’m in marketing, I have a post-grad degree in astronomy and I’ve loved it forever (just as a hobby, not professionally haha). When I joined Nouns, naming a space object was on my idea list. Thrilled you’re making it happen, and I’m all in to help however I can. This is so nounish!

1 Like

Thanks! So excited to meet other astronomy enthusiasts here at Nouns! @eusoujp

We’d love to proudly announce our current results on the project. You can also check the details at our GitHub repository: changsun20/HuntingAsteroid (


We propose implementing deep learning in the detection process of asteroids. Using data gathered from Hubble Space Telescope archive, we generate a dataset with around 3000 pictures. Then we tine-tune a pretrained model (previously trained on ImageNet) and design an automatic processing pipeline for the data.

What’s the idea

Thanks for the Hubble Asteroid Hunter citizen science project and the data they published, we can generate our labelled datasets for further image classification task.

This is our method to generate the training data:

We implement fine-tuning techniques to the classification task. This is how we build up the model:

Current Results

Up to now, we achieved a best accuracy of 92.24% and recall of 91.25% when fine-tuning ResNet50. We are now improving the performance of our model.

What’s next

We need your support, Nouns community!

We have built up our first version of pipeline with limited computation resources and relatively small datasets. To improve the performance of our model and boost the possibility of naming an asteroid, we really need the small grant and any other form of support from the community to build up larger datasets and run the model on better GPUs.

If you love our idea, feel free to contribute to our repository at HuntingAsteroid (, or email me at

Let’s name an asteroid, together!


  1. Sandor Kruk et al. Hubble Asteroid Hunter: I. Identifying asteroid trails in Hubble Space Telescope images. Astronomy & Astrophysics, 661:A85, May 2022.
  2. Kaiming He et al. Deep Residual Learning for Image Recognition, December 2015. Issue: arXiv:1512.03385 arXiv:1512.03385.
  3. Preeti Cowan et al. Towards Asteroid Detection in Microlensing Surveys with Deep Learning. Astronomy and Computing, 42:100693, January 2023. arXiv:2211.02239.
1 Like