o Ni)@sdZddlmZddlmZddlmZddlZddlZddlZddlZ ddl m m Z ddlmZdZdZdZd Zd Zd Zd Zgd ZdZdZedgdZGdddejjZddZ ddZ!ddZ"dS)Kitti dataset.)absolute_import)division)print_functionNa@inproceedings{Geiger2012CVPR, author = {Andreas Geiger and Philip Lenz and Raquel Urtasun}, title = {Are we ready for Autonomous Driving? The KITTI Vision Benchmark Suite}, booktitle = {Conference on Computer Vision and Pattern Recognition (CVPR)}, year = {2012} } aKitti contains a suite of vision tasks built using an autonomous driving platform. The full benchmark contains many tasks such as stereo, optical flow, visual odometry, etc. This dataset contains the object detection dataset, including the monocular images and bounding boxes. The dataset contains 7481 training images annotated with 3D bounding boxes. A full description of the annotations can be found in the readme of the object development kit readme on the Kitti homepage. z%http://www.cvlibs.net/datasets/kitti/z/https://s3.eu-central-1.amazonaws.com/avg-kittizdata_object_image_2.zipzdata_object_label_2.zipzdevkit_object.zip)CarVanTruck PedestrianPerson_sittingCyclistTramMisc RawBoundingBoxtopbottomleftrightc@sDeZdZdZejddZejdgZddZ ddZ d d Z d S) Kittirz3.2.0zDevkit updated.z3.1.0c Cstjjtdtjjdtjdtjjddtjjdtjdtjtjjdtjdtjjdtjdtjjdtjdd}tjj |t tj tj tj tj|dttd S) N)names)shapedtype) num_classes))type truncatedoccludedalphabbox dimensionslocation rotation_yimagezimage/file_nameobjects)builder descriptionfeatureshomepagecitation)tfdsr* ClassLabel_OBJECT_LABELSTensortffloat32 BBoxFeaturecore DatasetInfo _DESCRIPTION FeaturesDictImageTextSequence _HOMEPAGE_URL _CITATION)self annotationsrr^/home/ubuntu/.local/lib/python3.10/site-packages/tensorflow_datasets/object_detection/kitti.py_infoQs(    z Kitti._infoc Cstdttdttdtd}||}t||d\}}}tjj tj j ||d||dd|ddtjj tj j ||d||dd|ddtjj tj j ||d||dd|ddgS) N/)imagesr>devkitrCrBr>training)rBr>subdir image_ids)name gen_kwargs) _DATA_URL _IMAGES_FNAME _LABELS_FNAME _DEVKIT_FNAMEdownload _build_splits iter_archiver-r4SplitGeneratorSplitTRAIN VALIDATIONTEST)r= dl_manager filenamesfiles train_imagesvalidation_images test_imagesrrr?_split_generatorsisB            zKitti._split_generatorsccsRtjjj}t}|D])\}}tj|\} } | dkrq | tjj d|kr(q t ||t | dd<q |D]o\}}tj|\} } | dkrHq7| tjj d|krTq7t | dd} | |vraq7|| }| t j|t jd|j} || |j} | j\} }}|D]}t|d| ||d<|d=qtj|\}}| ||d }||fVq7dS) aYields images and annotations. Args: images: object that iterates over the archive of images. annotations: object that iterates over the archive of annotations. subdir: subdirectory from which to extract images and annotations, e.g. training or testing. image_ids: file ids for images in this split. Yields: A tuple containing the example's key, and the example. z.txtriNz.png)rbbox_rawr!r%)r-r4 lazy_importscv2dictospathsplitextsplitsep_parse_kitti_annotationsintimdecodenp fromstringreaduint8 IMREAD_COLORcvtColor COLOR_BGR2RGBr_build_bounding_box)r=rBr>rErFr^all_annotationsfpathfobjprefixextimage_idimgheightwidth_objfnamerecordrrr?_generate_exampless@      zKitti._generate_examplesN) __name__ __module__ __qualname____doc__r-r4VersionVERSIONSUPPORTED_VERSIONSr@r[r}rrrr?rIs  %rcCs4tjj||j|||j||j||j|dS)zBuilds and returns TFDS bounding box. Args: bbox: RawBoundingBox, bounding box in Kitti coordinates (origin top left). height: Image height in pixels. width: Image width in pixels. Returns: A TFDS BBox (origin bottom left). )yminymaxxminxmax)r-r*BBoxrrrr)r!rwrxrrr?ros   rocCsg}|D]X}ttj|gddd\}}}}}}} } } } } }}}}|dkr(q||t|t|t|tt|t| t|t| ddd| | | fDdd|||fDt|d q|S) zLoads and parses the Kitti object annotations. Args: annotations_csv: csv file containing annotations for a single image. Returns: A list of labelled bounding boxes. Each bounding box is stored as a dictionary of features.  ) delimiterrDontCarercSg|]}t|qSrfloat.0vrrr? z,_parse_kitti_annotations..cSrrrrrrr?rr)rrrr r\r"r#r$)listcsvreaderdecodeappendrrfr)annotations_csvr>lineobj_typerrr rrrrrwrxlengthxyzr$rrr?res0   recCsd}d}|D]3\}}|tjddkr#dd|ddD}q|tjddkr9|}d d|D}q|s>J|sBJtt }t |D]\}}||}|d d } ||  |qKt j } t j jd d td tt|d} td tt|d} tt j jtt || dd} tt j jtt t|t| | dd}g}g}g}|D]\}}|| vr||q||vr||q||qt j | |||fS)avSplits the train data into train/val/test by video. Ensures that images from the same video do not traverse the splits. Args: devkit: object that iterates over the devkit archive. Returns: train_images: File ids for the training set images. validation_images: File ids for the validation set images. test_images: File ids for the test set images. Nmappingztrain_rand.txtcSsg|] }t|dqS))rfstriprrrrr?rsz!_build_splits..utf-8,ztrain_mapping.txtcSsg|]}|dqS)r)rrrrr?r srr{)seeddF)replace)r`rajoinrjrrc splitlines collections defaultdictr enumeraterrhrandom get_statermax_TEST_SPLIT_PERCENT_VIDEOSlen _VALIDATION_SPLIT_PERCENT_VIDEOSsetchoicesortedkeysitemsextend set_state)rCmapping_line_ids mapping_linesrqrrvideo_to_imagerumapping_lineidrvideo_idnumpy_original_statenum_test_videosnum_validation_videos test_videosvalidation_videosrZrYrXkrrrr?rNsd          rN)#r __future__rrrrrr`numpyrhtensorflow.compat.v2compatv2r1tensorflow_datasets.public_api public_apir-r<r6r;rIrJrKrLr/rr namedtuplerr4GeneratorBasedBuilderrrorerNrrrr?s6      u #