Bees vs Wasps

Synopsis

Hand-curated, close-up photos of bees, wasps, and other insects. The challenge is primarily to distinguish bees from wasps.

example bees:

example wasps:

example other insect:

example other non-insect:

Excerpt from labels.csv :

 id,path,is_bee,is_wasp,is_otherinsect,is_other,photo_quality,is_validation,is_final_validation 1,bee1\10007154554_026417cfd0_n.jpg,1,0,0,0,1,0,0 2,bee1\10024864894_6dc54d4b34_n.jpg,1,0,0,0,1,0,1 3,bee1\10092043833_7306dfd1f0_n.jpg,1,0,0,0,1,1,0 6842,wasp2\I00101.jpg,0,1,0,0,0,0,0 6843,wasp2\I00102.jpg,0,1,0,0,0,0,0 6844,wasp2\I00103.jpg,0,1,0,0,0,1,0 

Dataset totals

we have:  bees..........: 3183  wasps.........: 4943  other insects.: 2453  other.........: 845  in that, there is:  training photos : 7942  hyperparameter tuning (1st level validation) photos : 1719  final validation (brag about your result with these) photos : 1763  In the final validation, there is 504 bees and 753 wasps, meaning that the resolution of the result is 0.08% 

Labels:

in labels.csv :

• id - ordinal - unique index
• path - string - relative path to the photo, including extension
• is_bee - nominal - 1 if there is a bee in the photo
• is_wasp - nominal - 1 if there is a wasp in the photo
• is_otherinsect - nominal - 1 if there is other insect prominently in the centre of the photo, but it is not a wasp and not a bee. It might be a fly, but there are other things there too, like beetles
• is_other - random photos not containing any insects
• photo_quality - 1 for photos where I have very high confidence that it is bee, wasp, or other. 0 for photos of generally low quality or where I am not very confident that it is what it says it is. You can use this to initially reduce the size of the training set
• is_validation - you can use this for your training validation, or you can combine these with the training data and split your training/validation differently
• is_final_validation - do NOT use these photos for training - use them to compute your final score. This will enable comparing results by different kagglers. Optionally, if you want to deploy an app to actually serve the model, you can then use these for final training too.

Credits

This image dataset collates and refines upon several sources:

• "PollenDataset" by 2017 Ivan Rodriguez, Rémi Mégret, Edgar Acuña, José Agosto, Tugrul Giray, from https://www.kaggle.com/ivanfel/honey-bee-pollen

• https://www.kaggle.com/tegwyntwmffat/european-wasp-vespula-vulgaris-kitti-format/ - there isn't any better description

• Flicker search for bee, wasp and fly

The photos have been hand-curated by our expert biologist, Callum Robertson https://www.linkedin.com/in/callum-robertson-358014109/

Collator and Kaggle competitor: George Rey https://www.linkedin.com/in/dr-george-rey-dziewierz/

​

---

Contributions:

Ivan Rodriguez, Rémi Mégret, Edgar Acuña, José Agosto, Tugrul Giray. Recognition of pollen-bearing bees from Video using Convolutional Neural Network, IEEE Winter Conf. on Applications of Computer Vision, 2018, Lake Tahoe, NV. https://doi.org/10.1109/WACV.2018.00041

Inspiration

Some notes:

• The data is probably biased:
  • Most bees are photographed on a flower, while most wasps are not. Your AI might learn to recognize the existence of a flower rather than existence of a bee.existence of a bee. You might want to test it against some "empty flowers" to verify.