Skip to content
Amazon Picking Challenge: Recognition
C++ C CMake
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.

#Amazon Picking Challenge

How to install EBlearn

The eblearn package is in uts_recogniser/externals/. The installation instruction is here. Both eblearn core and tools are required to be installed. EBlearn does NOT provide default link file. If EBlearn is installed correct to /usr/include and /usr/lib, eblearn.pc is able to find EBlearn on your PC.

Please check the EBlearn is installed and tested successfully before compling the updated uts_recogniser.

How to run eblearn test?

.../bin/eblearn_test ../data/eblearn/images/kong_duck_dog_toy.bmp ../data/eblearn/ kygen_squeakin_eggs_plush_puppies kong_duck_dog_toy

How to run it?

Normally, I am more used to run ros package directly using binray file instead of rosrun <package name> <binary name>.

  1. Test Kernel Descriptor at first:

    ../bin/kd_test ../data/201504174639/xtion_rgb_1.png ../data/201504174639/bin_A_empty.png ../data/201504174639/mask_xtion_rgb_bin_A.png feline_greenies_dental_treats mead_index_cards expo_dry_erase_board_eraser kong_duck_dog_toy

  2. Test offline recogniser:

    ../bin/offline_recogniser -j ../data/201504171107.json -mask ../data/201504174639/ -method ../data/method.txt -kd ../data/kd_models/

    ../bin/pseudo_request -d ../data/201504174639/ -n 12 -j ../data/201504171107.json

Object recognition status:

* R = Recognisable, correct
* NR = Not Recognisable, will be recognised using Machine Learning
* NA = Not available, cannot be purchased in Australia
* O = Ordered, no available yet
Object name Status
oreo_mega_stuf NA
champion_copper_plus_spark_plug R
expo_dry_erase_board_eraser R
genuine_joe_plastic_stir_sticks R
munchkin_white_hot_duck_bath_toy NR
crayola_64_ct R
mommys_helper_outlet_plugs NR
sharpie_accent_tank_style_highlighters NR/R1
stanley_66_052 NR
safety_works_safety_glasses NR
cheezit_big_original R
paper_mate_12_count_mirado_black_warrior R
feline_greenies_dental_treats R2
elmers_washable_no_run_school_glue R3
mead_index_cards R
rolodex_jumbo_pencil_cup NR
first_years_take_and_toss_straw_cup NR
highland_6539_self_stick_notes R
mark_twain_huckleberry_finn R
kyjen_squeakin_eggs_plush_puppies NR
kong_sitting_frog_dog_toy NR
kong_air_god_squeakair_tennis_ball NR
dr_browns_bottle_brush NR/R4
kong_duck_dog_toy NR
laugh_out_loud_joke_book R

Detailed explanations:

1. *sharpie_accent_tank_style_highlighters* can be recognised only using the front view, using the label in the front.
2. *feline_greenies_dental_treats* is non-rigid, however, this item can be recognised only if we don't bend the item too much. **model updated**
3. *elmers_washable_no_run_school_glue* can be recognised, the model will be updated lated this week. **model updated**
4. *dr_browns_bottle_brush* can roughly recognised using the texture on the paper board
You can’t perform that action at this time.