Skip to content

Instantly share code, notes, and snippets.

View lucasb-eyer's full-sized avatar
🤔
{D,R}L Research @ Google Brain Zürich

Lucas Beyer lucasb-eyer

🤔
{D,R}L Research @ Google Brain Zürich
View GitHub Profile
@lucasb-eyer
lucasb-eyer / instructions.md
Created April 13, 2026 06:43
Prompt to make IB tws work on wayland arch linux

IBKR Desktop Wayland Fix Notes

This install is in the current workdir.

What broke

IBKR Desktop uses Qt Jambi and is sensitive to the native Qt version it loads. The app's bundled Qt was 6.8.3, but at one point some files in lib/qt/lib/ and lib/qt/plugins/platforms/ were replaced with Arch system Qt 6.11 pieces. That caused the launcher to fail early with a Qt Jambi native mismatch.

We can make this file beautiful and searchable if this error is corrected: It looks like row 2 should actually have 1 column, instead of 3 in line 1.
TsvHttpData-1.0
https://image-net.org/data/winter21_whole.tar.gz 1185381173159 qzE84DF5/YA6QBsCxlHAog==
@lucasb-eyer
lucasb-eyer / main.py
Created July 28, 2020 19:43
Puzzle falling onto table
# Commands can be sent in lists of arbitrary length, allowing for arbitrarily complex instructions per frame. The user must explicitly request output data:
from tdw.controller import Controller
from tdw.tdw_utils import TDWUtils
from tdw.librarian import ModelLibrarian
from tdw.output_data import OutputData, Bounds, Images
c = Controller(launch_build=True)
lib = ModelLibrarian("models_core.json")
@lucasb-eyer
lucasb-eyer / lb_prec_rec.py
Last active February 10, 2018 11:12
Python code for 2D precision-recall computation of point-detections. Audited by @Pandoro :)
from collections import defaultdict
import numpy as np
from scipy.optimize import linear_sum_assignment
from scipy.spatial.distance import cdist
from sklearn.metrics import auc
def prec_rec_2d(det_scores, det_coords, det_frames, gt_coords, gt_frames, gt_radii):
""" Computes full precision-recall curves at all possible thresholds.
Arguments:
@lucasb-eyer
lucasb-eyer / count_cnn_eccv.sh
Created October 12, 2016 00:24
Count approximate number of deep-learning/cnn posters in ECCV
# count number of posters
wget -O - http://www.eccv2016.org/main-conference/ | grep -i 'pdf_icon' | wc -l
# == 504
# number of deep-learning ones
wget -O - http://www.eccv2016.org/main-conference/ | grep -i 'deep\|convolutional\|network\|cnn\|neural' | wc -l
# == 127
# Note the numbers are approximates because some posters are present twice, there's a legend on the icons, etc.
class Dummy():
def __init__(self, *a, **kw):
pass
def __setstate__(self, *a, **kw):
pass
def __getstate__(self, *a, **kw):
pass
@lucasb-eyer
lucasb-eyer / ILSVRC2012_ids.json
Last active June 10, 2017 02:37
Data for Beacon8 Pretrained Imagenet example
{
"n02112350": 148,
"n04344873": 311,
"n01692333": 470,
"n03459775": 725,
"n04133789": 751,
"n01871265": 214,
"n04366367": 681,
"n03891332": 527,
"n03085013": 543,
@lucasb-eyer
lucasb-eyer / gist:ace20ce1487063947614
Created February 7, 2015 12:25
awesomewm battery widget
-- Create a battery widget on laptops
mybattery = nil
mybattery_timer = nil
if lucasb.file_exists("/sys/class/power_supply/BAT0/status") then
mybattery = awful.widget.progressbar()
mybattery:set_width(13)
mybattery:set_vertical(true)
mybattery_tip = awful.tooltip({objects = {mybattery}})
lastalarm = 0
@lucasb-eyer
lucasb-eyer / test.cpp
Last active August 29, 2015 14:05
Testing JPG -> MagickWand -> raw buffer -> MagickWand -> JPG
#include <wand/magick_wand.h>
#include <iostream>
#include <vector>
using namespace std;
int main(int argc, char* argv[])
{
MagickWand *mw_in = NewMagickWand();
MagickWand *mw_out = NewMagickWand();
@lucasb-eyer
lucasb-eyer / output
Created January 19, 2014 00:14
SVM.jl prediction performance
Fitted linear SVM
* Non-zero weights: 100
* Accuracy: 0.963000
* Iterations: 100
* Converged: true
Computing Raw scores
--------------------
vectorized julia
elapsed time: 0.641106641 seconds (80048000 bytes allocated)