![\"Shipping\"](\"https://storage.googleapis.com/kaggle-media/competitions/Airbus/ships.jpg\")
Airbus is excited to challenge Kagglers to build a model that detects all ships in satellite images as quickly as possible. Can you find them even in imagery with clouds or haze?
\n \nHere’s the backstory: Shipping traffic is growing fast. \nMore ships increase the chances of infractions at sea like environmentally devastating ship accidents, piracy, illegal fishing, drug trafficking, and illegal cargo movement. This has compelled many organizations, from environmental protection agencies to insurance companies and national government authorities, to have a closer watch over the open seas.
\n\n
Airbus offers comprehensive maritime monitoring services by building a meaningful solution for wide coverage, fine details, intensive monitoring, premium reactivity and interpretation response.\nCombining its proprietary-data with highly-trained analysts, they help to support the maritime industry to increase knowledge, anticipate threats, trigger alerts, and improve efficiency at sea.
\nA lot of work has been done over the last 10 years to automatically extract objects from satellite images with significative results but no effective operational effects. Now Airbus is turning to Kagglers to increase the accuracy and speed of automatic ship detection.
\n\nAlgorithm Speed Prize: After the Kaggle challenge is complete, competitors may submit their model via a private Kaggle kernel for a speed evaluation based upon the inference time on over 40.000 images chips (typical size of a full satellite image) to win a special algorithm speed prize.
\n\n\nIf you're interested to explore more Airbus data, you are welcomed to check out the OneAtlas Sandbox. And for more insights on our Maritime Surveillance capabilities, have a look at Airbus Intelligence page.", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Evaluation", "content": "This competition is evaluated on the F2 Score at different intersection over union (IoU) thresholds. The IoU of a proposed set of object pixels and a set of true object pixels is calculated as:\n$$\nIoU(A,B) = \\frac{A \\cap B}{ A \\cup B}.\n$$\nThe metric sweeps over a range of IoU thresholds, at each point calculating an F2 Score. The threshold values range from 0.5 to 0.95 with a step size of 0.05: `(0.5, 0.55, 0.6, 0.65, 0.7, 0.75, 0.8, 0.85, 0.9, 0.95)`. In other words, at a threshold of 0.5, a predicted object is considered a \"hit\" if its intersection over union with a ground truth object is greater than 0.5.\n\nAt each threshold value \\\\(t\\\\), the F2 Score value is calculated based on the number of true positives (TP), false negatives (FN), and false positives (FP) resulting from comparing the predicted object to all ground truth objects. The following equation is equivalent to F2 Score when \\\\( \\beta \\\\) is set to 2:\n\n$$\nF_\\beta(t) = \\frac{(1 + \\beta^2) \\cdot TP(t)}{(1 + \\beta^2) \\cdot TP(t) + \\beta^2 \\cdot FN(t) + FP(t)}.\n$$\n\nA true positive is counted when a single predicted object matches a ground truth object with an IoU above the threshold. A false positive indicates a predicted object had no associated ground truth object. A false negative indicates a ground truth object had no associated predicted object. The average F2 Score of a single image is then calculated as the mean of the above F2 Score values at each IoU threshold:\n\n$$\n\\frac{1}{|thresholds|} \\sum_t F_2(t).\n$$\n\nLastly, the score returned by the competition metric is the mean taken over the individual average F2 Scores of each image in the test dataset.\n\n## Kernel Submissions\nYou can make submissions directly from Kaggle Kernels. By adding your teammates as collaborators on a kernel, you can share and edit code privately with them.\n\n## Submission File\nIn order to reduce the submission file size, our metric uses run-length encoding on the pixel values. Instead of submitting an exhaustive list of indices for your segmentation, you will submit pairs of values that contain a start position and a run length. E.g. '1 3' implies starting at pixel 1 and running a total of 3 pixels (1,2,3).\n\nThe competition format requires a space delimited list of pairs. For example, '1 3 10 5' implies pixels 1,2,3,10,11,12,13,14 are to be included in the mask. The pixels are one-indexed \nand numbered from top to bottom, then left to right: 1 is pixel (1,1), 2 is pixel (2,1), etc. A prediction of of \"no ship in image\" should have a blank value in the `EncodedPixels` column.\n\nThe metric checks that the pairs are sorted, positive, and the decoded pixel values are not duplicated. It also checks that no two predicted masks for the same image are overlapping.\n\nThe file should contain a header and have the following format. Each row in your submission represents a single predicted ship segmentation for the given image.\n\n ImageId,EncodedPixels\n 00002bd58.jpg,1 3\n 00015efb6.jpg,\n 00023d5fc.jpg,1 3 10 5\n etc.\n", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Prizes", "content": "
The goal of the Algorithm Speed Prize is to encourage participants to develop the fastest running algorithm that will still produce accurate results. The speed prize will be open to the top 100 performing teams on the private leaderboard.
\nIf you want to participate in the algorithm speed prize, you will need to use Kaggle Kernels and take the steps outlined in the Algorithm Speed Prize tab
\n\n\n**Timeline**\n\n* **November 30, 2018** 11:59 PM UTC - Final submission deadline to share the private kernel with host.\n* **December 14, 2018** - Announcement of the winner of the Algorithm Speed Prize.", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Timeline", "content": "**New Competition End Dates**\n\n* **November 7, 2018** - Entry deadline. You must accept the competition rules before this date in order to compete.\n\n* **November 7, 2018** - Team Merger deadline. This is the last day participants may join or merge teams.\n\n* **November 14, 2018** - Final submission deadline.\n\n* **November 30, 2018** - Final submission deadline for the Algorithm Speed Prize.\n* **December 14, 2018** - Announcement of the winner of the Algorithm Speed Prize.\n\nAll deadlines are at 11:59 PM UTC on the corresponding day unless otherwise noted. The competition organizers reserve the right to update the contest timeline if they deem it necessary.", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Algorithm Speed Prize ", "content": "**Background**\n\nTo fight against unwelcome infractions at sea like environmentally devastating ship accidents, piracy, illegal fishing or illegal cargo movement, we need to act fast. Airbus strives to go under a few tens of minutes from image acquisition on our satellites to processing in our data centres. So processing time has to stay low! That is why Airbus is offering an algorithm speed prize to the fastest running algorithm that will still produce very good results!\n\n**Evaluation**\n\nThe speed prize will be open to teams that score in top 100 of the private leaderboard. If you want to participate in the algorithm speed prize, you will need to follow the steps below. \n\nSteps to Enter:\n\n 1. Create a private dataset that includes model weights and any other\n elements required to make predictions.\n 2. Create a private kernel that includes the code to make predictions on the test set. (Note: If you win the speed prize, you will also be required to submit the code that generated the model.)\n 3. Run the kernel, making note of the inference time. [see code below]\n 4. Make a post-competition submission to the leaderboard and note the score.\n 5. **You should add the inference time and the private leaderboard score of your kernel as a discussion comment at the bottom of the kernel.**\n 6. Add Jeff Faudi (jeffaudi) and Walter Reade (inversion) as collaborators to both the Kernel and Dataset.\n 7. The Kernel will be validated against a new dataset.\n\nThe winner will be the fastest Kernel (that does not have significant degradation in predictive ability against the validation set).\n\nTo report inference time:\n\n import time\n inference_start = time.time()\n >> inference code <<\n inference_end = time.time()\n print('Inference Time: %0.2f Minutes'%((inference_end - inference_start)/60))\n\n\n**Timeline**\n\n* **November 30, 2018** - Final submission deadline for speed prize kernel submission.\n* **December 14, 2018** - Announcement of the winner of the Algorithm Speed Prize.\n\nAll deadlines are at 11:59 PM UTC on the corresponding day unless otherwise noted. The competition organizers reserve the right to update the contest timeline if they deem it necessary.", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false } ] }