Accelerating your geospatial deep learning pipeline with fine-tuning

Using CosmiQ’s Solaris library to fine-tune models pre-trained on SpaceNet overhead imagery

Nick Weir
Nick Weir
Sep 11, 2019 · 4 min read

Summary: Using Solaris, you can fine-tune deep learning models pre-trained on overhead imagery for five minutes and achieve performance comparable to past SpaceNet Challenge prize-winners.

The 5th round of the popular SpaceNet Challenge series has begun! There, participants are challenged to develop models that can extract routable road networks with travel time predictions from overhead imagery. Past iterations of the challenge have been dominated by deep learning models painstakingly trained for days, often using multiple GPUs. The winner in the most recent SpaceNet challenge ensembled 28 independently trained models to produce their solution, which entailed ~650 GPU-hours of training on commercial-grade GPUs — making it hard for participants with limited computing resources to be competitive. We at CosmiQ Works have frequently asked ourselves: how can we make the SpaceNet Challenges more accessible to individuals who may lack these resources?

Source image (left), hand-labeled buildings (middle), and buildings predicted by XD_XD’s model trained on Atlanta (right). The model does very poorly when it has never “seen” any imagery of Khartoum before.
A comparison of the original image of Khartoum (top left), the hand-labeled ground truth (bottom left), the predicted buildings before fine-tuning (top right), and the predictions after fine-tuning (bottom right). Though still imperfect, the buildings identified after fine-tuning are markedly better than before fine-tuning. This image was held out from the training set during fine-tuning.

The DownLinQ

Welcome to the archived blog of CosmiQ Works, an IQT Lab

The DownLinQ

As of March 2021, CosmiQ Works has been folded into IQT Labs. An archive will remain here to showcase historical work from CosmiQ Works that took place July 2016 — March 2021.

Nick Weir

Written by

Nick Weir

Data Scientist at CosmiQ Works and SpaceNet 4 Challenge Director at the SpaceNet LLC. Advancing computer vision and ML analysis of geospatial imagery.

The DownLinQ

As of March 2021, CosmiQ Works has been folded into IQT Labs. An archive will remain here to showcase historical work from CosmiQ Works that took place July 2016 — March 2021.