eprintid: 10155879 rev_number: 10 eprint_status: archive userid: 699 dir: disk0/10/15/58/79 datestamp: 2022-09-21 11:59:32 lastmod: 2022-09-21 11:59:33 status_changed: 2022-09-21 11:59:32 type: proceedings_section metadata_visibility: show sword_depositor: 699 creators_name: Wang, H creators_name: Liu, Q creators_name: Yue, X creators_name: Lasenby, J creators_name: Kusner, MJ title: Unsupervised Point Cloud Pre-training via Occlusion Completion ispublished: pub divisions: C05 divisions: F48 divisions: B04 divisions: UCL note: This version is the author accepted manuscript. For information on re-use, please refer to the publisher's terms and conditions. abstract: We describe a simple pre-training approach for point clouds. It works in three steps: 1. Mask all points occluded in a camera view; 2. Learn an encoder-decoder model to reconstruct the occluded points; 3. Use the encoder weights as initialisation for downstream point cloud tasks. We find that even when we pre-train on a single dataset (ModelNet40), this method improves accuracy across different datasets and encoders, on a wide range of downstream tasks. Specifically, we show that our method outperforms previous pre-training methods in object classification, and both part-based and semantic segmentation tasks. We study the pre-trained features and find that they lead to wide downstream minima, have high transformation invariance, and have activations that are highly correlated with part labels. Code and data are available at: https://github.com/hansen7/OcCo. date: 2021 date_type: published publisher: Institute of Electrical and Electronics Engineers (IEEE) official_url: https://doi.org/10.1109/ICCV48922.2021.00964 oa_status: green full_text_type: other language: eng primo: open primo_central: open_green verified: verified_manual elements_id: 1821281 doi: 10.1109/ICCV48922.2021.00964 isbn_13: 978-1-6654-2812-5 lyricists_name: Kusner, Matthew lyricists_id: MKKUS92 actors_name: Flynn, Bernadette actors_id: BFFLY94 actors_role: owner full_text_status: public pres_type: paper publication: Proceedings of the IEEE International Conference on Computer Vision pagerange: 9762-9772 event_title: 2021 IEEE/CVF International Conference on Computer Vision (ICCV) event_location: Montreal, QC, Canada event_dates: 11th-17th October 2021 book_title: Proceedings of the IEEE International Conference on Computer Vision citation: Wang, H; Liu, Q; Yue, X; Lasenby, J; Kusner, MJ; (2021) Unsupervised Point Cloud Pre-training via Occlusion Completion. In: Proceedings of the IEEE International Conference on Computer Vision. (pp. pp. 9762-9772). Institute of Electrical and Electronics Engineers (IEEE) Green open access document_url: https://discovery.ucl.ac.uk/id/eprint/10155879/1/Wang_Unsupervised_Point_Cloud_Pre-Training_via_Occlusion_Completion_ICCV_2021_paper.pdf