F. Güney
@ftm_guney
Followers
8,451
Following
1,410
Media
467
Statuses
6,175
research on computer vision, teaching, and movies. asst. prof. @KuisAICenter @kocuniversity tweets in TR, EN
Istanbul, Turkey
Joined July 2009
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
JENNIE FOR CHANEL
• 292569 Tweets
Nigeria
• 231224 Tweets
Happy New Month
• 137310 Tweets
Iniesta
• 43386 Tweets
#格付けチェック
• 42580 Tweets
레이예스
• 40322 Tweets
メガネの日
• 35441 Tweets
#大橋和也
• 27490 Tweets
対象作品
• 19003 Tweets
集英社秋マンガチャ開催
• 17081 Tweets
日本酒の日
• 13577 Tweets
#千燈ゆうひ3D
• 13333 Tweets
#優しく強くアヌア
• 13017 Tweets
Pinned Tweet
check out our recent work CarFormer, to be presented at
#ECCV
:
given BEV sequences*, we extract objects into slots and feed them as tokens to an autoregressive transformer to learn how to drive and predict future slot representations.
operating at
5
6
67
I'm teaching Computer Vision with Deep Learning this semester. any example courses that you can recommend to check for reference (other than CS231n obviously)?
45
82
1K
it still feels unreal but we can finally talk about it: I've received an ERC Starting Grant with my project ENSURE.
with this funding by
@ERC_Research
, I will explore different types of uncertainty in self-driving.
I'm extremely grateful for all the support 😇
#ERCStG
68
3
408
I still don’t have a visa for ICCV. let’s keep telling ourselves how inclusive we are.
the summer I got 3 papers and an ERC and still wasn’t allowed in Europe.
if you read this and think I deserve this because I want to live in the same country as my family, just unfollow me.
5
28
406
I got the Marie Curie Individual Fellowship!!
dancing with my mom and dad at the moment:
26
3
401
don’t invite me to “a virtual event to recognize the forward momentum and impact of Women blabla” give me funding, give me GPUs, accept my students to your internship programs, cite my papers. I don’t need to be told that I’m doing great at a virtual event, I need actual support.
4
25
390
I’m losing my best students to stupid app companies because of money. how is that good for anyone.. I’m not even angry, I just feel tired and maybe a little hopeless. time for kickbox, you also have to learn how to take a hit.
10
10
384
some NeurIPS highlights:
LDM tutorial was great. the 1st part is a must-watch, the 2nd part is full of references for recent advancements, and you need to check the 3rd part if you're planning to use LDMs for something other than images/videos.
4
45
292
my visa application got rejected again. for the third time. in a single summer. tell me this is not political. included my a**
21
14
283
yayy I got an outstanding research award at the uni today 🥳
many thanks to our president Metin Hoca and our dean Attila Hoca, not only for the nomination but also for the support 😇
27
3
286
Richard Hartley finding a bug in the formulation of NeRF that makes it easier to optimize.. this day has already turned out more fun than I thought.
5
21
276
Jitendra Malik: "Computer Vision is too important to be left to just Machine Learning people."
mic drop moment?
#ICCV2021
I love that he calls it an "awareness" as a whole 😍
3
34
261
I cannot believe we did not get rejected this time 🥹 actually, someone check the acceptance ratio because we did 3 in 3 🔥 I’m so happy for my students 😍 I cannot think of a better a birthday present --actually I can but no pressure ERC 😅
14
0
219
optimistic Americans using reward functions vs. pessimistic Russians using cost functions 😂 by
@svlevine
3
15
206
yayy
@ICCV_2021
🥳 congrats
@akaan_akan
! our first paper as a group, this is a monumental moment!
19
3
202
@oktayarslan
Oktay Bey, sonradan eklemeler yaptım, onları da kopyalayın:
bu süreçte yaşayacak kadar kazanmak lazım tabi ama para öncelik olmasın. either earn or learn. Tübitak bursu hiç yoktan iyidir, undergradlar için STAR programı var. biraz da reklam: KUIS AI olarak en yüksek bursları veriyoruz. yurt olmazsa yine çok yetmiyor ama onu da çözeceğiz.
1
1
13
7
2
200
it's very discouraging to get "is this included in the exam?" when you go above and beyond to teach a difficult topic.
18
3
191
great news!! we got the Newton Advanced Fellowship with our proposal "The Road Less Travelled: One-Shot Learning of Rare Events in Autonomous Driving" together with João and Luca from
@Oxford_VGG
! £111K for 3years.
@often_worried
21
4
185
we’re on the news!
many thanks to Ralph for his unwavering support 😊
Good morning ICCV 2023!
Hot off the presses, ICCV Daily News:
What happened and what is going to happen at ICCV2023 in Paris today. Enjoy!
#ICCV
#ICCV2023
#ComputerVision
#CVPR
@ICCVConference
It's a publication by
@RSIPvision
and Computer Vision News!
0
2
27
1
19
186
undergrad olarak AI ile ilgileniyorsak hangi dersleri alalım, nasıl bir yol izleyelim diye çok soru geliyor. öncelikle sağlam bir temel için calculus, linear algebra ve probability çok iyi öğrenmeniz lazım. temel sağlam olmazsa sonra toparlaması zor oluyor, zaman alıyor gereksiz.
4
23
179
reddedilen Tübitak projemin raporu geldi. full trajectory prediction algoritması “sadece şerit değiştirme” diye eleştirilmiş, ülkenin sosyo-ekonomik hedeflerine uymuyormuş 🙀
zaten projede çalışacak dereceli iki öğrenci gitti, mobile game geliştiriyor.
ne ülke ama, ne gelecek 👌🏾
1
3
177
Our survey "Computer Vision for Autonomous Vehicles: Problems, Datasets and State of the Art" is now published as a book in FnT CGV.
AS-A-BOOK!!💃
4
21
179
our new work “A Likelihood Ratio-Based Approach to Segmenting Unknown Objects” is on arXiv:
we augment a pre-trained inlier segmentation network with an unknown estimation module to estimate a proxy outlier distribution and use it to calculate a
2
23
179
once, I told my grandmother I'm working on teaching vehicles to drive themselves. her reaction was: "but you don't know how to drive yourself!"
no better review has been said or written.
4
5
172
why does mixup work so well for OoD? but why always a little mixed up, e.g. 50-50 mixup will not give you good results? it's because the model never sees regular in-distribution samples. simply add a loss on your regular samples as well. amazing finding!
3
26
161
we can now talk about “RbA: Segmenting Unknown Regions Rejected by All” accepted at
#ICCV2023
.
we found that when we find and classify regions with Mask2Former, each object query focuses on a known class. so we can find unknowns as pixels rejected by all, hence the funny name 😃
do you think a flying car is OoD? our algorithm certainly thinks so:
(also OMG 😱)
1
4
38
6
19
161
got rejected from some google research award thing, again. sending this tweet partly because I cannot convince myself that I would not send it if accepted but also to normalize rejection and convince desperados like myself to keep trying.
5
2
158
Laura and Torsten are organizing a tutorial on "How to write a good review?" at
#CVPR20
on a mission to
#makeReviewsGreatAgain
🙂
It is truly an honor to be invited to this workshop alongside great speakers.
5
28
141
I’m teaching an Intro to programming and CS course, do you know how many *female* students I had this first week who came and told me that they don’t know any programming, they are not sure if they can do it, if CS is right for them? I’ve stopped counting.
6
5
134
Kaan successfully defended his Master’s thesis today! it was an amazing presentation with great questions and suggestions from the jury members, Aykut Erdem and Jurgen Gall. it was a very happy and proud moment because Kaan is the first student who graduated from our group 😊
3
1
128
we couldn’t attend CVPR -no paper/money/visa- so we got together today to watch some talks. it’s great that most of them are already on YouTube.
it was a lot of fun, we even played foosball during breaks!
special thanks to
@inzvaspace
for hosting us in their new awesome place!
1
3
129
of course, turn to the man next to me while asking "who is buying the car?" well I guess you get your patriarchy in check while some dreams are coming true for me 😊
8
0
119
I’m presenting our paper with
@gorkaydemir
and
@WeidiXie
tomorrow at Poster Session 2* at
#NeurIPS2023
:
with SOLV; “Self-supervised Object Centric Learning for Videos”, we can discover multiple objects in real-world video sequences without using additional modalities like depth
1
15
123
in case you're looking for role models in Computer Vision, here is three of mine 🙂 rocking it 🤘🏽💃🏽
1
11
116
Jitendra: “vision’s first love was robotics. then we had an affair first with graphics, then with machine learning. it’s time to go back to our first love.” 💕
1
8
116
We're looking for highly motivated graduate students for our project "The Road Less Travelled: One-Shot Learning of Rare Events in Autonomous Driving", also in collaboration with João Henriques (
@Oxford_VGG
) and
@often_worried
(
@_FiveAI
).
4
23
116
our workshop proposal on “what is motion for?” has been accepted at
#ECCV2022
! after the success of the first one at ECCV2018, with a bigger team this time, we’re looking forward to investigating what the community thinks about motion!
6
7
115
CVPR reviews arrive today 🥹 how do you cope with this? after almost 10 years, I’m still terrified to wake up in the middle of the night (US time zones) and read hurtful things three strangers say about our precious 🥹😂
I tried to give feedback to reviewers in my batch to tune
5
1
115
I’m walking the
#LycianWay
this week with some amazing views in the south west of Turkey. finding my way so far despite the first picture where I go in the opposite direction at the beginning 😅
1
1
106
we’ve got multiple PhD and postdoc positions funded by my
#ERCstg
project ENSURE.
if you’re interested in computer vision and self-driving, please consider applying.
graduate students: apply ASAP! details at
postdocs: send me an email with your CV and
8
31
109
We've revised our survey paper
"Computer Vision for Autonomous Vehicles: Problems, Datasets and State of the Art"
Lots of new papers, better organization and more insights!
(to appear as a book in Foundations and Trends in Computer Graphics and Vision)
4
24
107
yayy I've just accepted to give a talk at WiCV this CVPR! I'm thrilled and terrified at the same time because I have no idea what I'm going to talk about, and my heart rate is close to when I'm running but heyy do the things that scare you the most, right 😅
3
0
108
R2: what is the selling point of this paper?
me: I'm so glad you asked. we wrote an 8-page document on this but let's give it another try to convince you with one more page.
rebuttal feels pointless sometimes.
6
5
106
can visual foundation models achieve long-term point tracking? (spoiler alert: yes, they can!)
check out the detailed analysis by
@gorkaydemir
, co-supervised by amazing
@WeidiXie
:
to be presented at the EVAL-FoMo workshop this
#ECCV2024
.
1
17
107
Tübitak sağolsun ERC training için destek oldu. Gönderdikleri para TL, ödeme Euro olunca, aynı gün çevirmeme rağmen 4-5 bin TL içerdeyim.
o kadarını da kendim ödeyivereyim tamam da, neden bunları yaşıyoruz biz? sinirli falan değilim, genel çok üzgünüm. sürekli üzgünüm bu ülkede.
3
2
102
I got the grant to start my own group!
2232 olumlu, inanıyorum, her şey çok güzel olacak!
15
1
103
I'm probably the most annoying AC ever 🤓
- Which parts of the method do you think need more details?
- What kind of ablation studies/datasets specifically do you recommend?
- Could you add some references to verify your claim on "limited novelty" and point to the similarities?
6
4
103
my worldview about the LLMs can be found attached:
3
15
98
come on ICCV, reject us already!! who else is afraid of clicking on “Withdraw submission” while refreshing CMT 🥹
7
1
96
I’ve seen some advice on networking which recommends chasing bigshots instead of hanging out with your peers, especially from lower ranking institutes, e.g. at grad student parties. it’s complete BS. those ppl you hang out with will become a prof or researcher somewhere and they
5
4
94
this year, in course projects, I asked each team to choose a SOTA model and make it fail in interesting ways. one group chose "Tracking through Containers and Occluders in the Wild"* and wild it got 😁
the poor model did not even see what hit it 😄
*
5
8
92
meet the team! minus Sadra and some of our interns.
left-to-right: Zeynep, Nursena, Ege, me, Taher, Misra, Nazir, Cagan, Caner, and Kaan.
2
0
89
no better way to celebrate women’s day than submitting a paper with a female student as the first author. I’m sorry but it’s a big deal, it doesn’t happen very often. in fact, the first in 4.5 years! let’s hope, first of many 🤞🏽
4
2
89
it was an amazing experience to speak at the
#CVPR20
tutorial on reviewing, despite my trembling voice and shaking hands 😅 huge thx to Luara and Torsten for giving me the chance, and to the speakers for great, fun talks. the internal comments made me LoL constantly on zoom 😂
10
1
86
Bilim Akademisi - Bilkent Üniversitesi Yapay Öğrenme Yaz Okulu'nda konuşmamı tamamladım.
150 kişi olarak başlayan konuşmanın sonunda 500 kişi olduğunu görünce, sunum sırasında zoom paneline bakmamanın ne kadar iyi bir fikir olduğunu görmüş olduk 😅
2
0
84
we got rejected because we improved the results in the rebuttal and apparently, it'd be unfair to accept it. I mean, sure. also, let's ignore that only 1 out of 3 reviewers entered their final rating.
11
7
87
our paper which addresses moving objects in monocular depth estimation got accepted to 3DV, it’s Sadra’s first paper 🥳 we’ve been working on this for a while, looking forward to sharing our results.
4
2
84
Özel üniversitelerde ancak 15 sene sonra yeşil pasaport alınabilmesi konusunu konuşabilir miyiz? 6 ayda bir Schengen başvurusu yapmaktan, bu yüzden bir şeyler kaçırmaktan bıktım. Randevu bulunamıyor doğru dürüst, bulsan bile eften püften sebeplerle reddedilebiliyor.
@mfatihkacir
3
7
84
my students are writing papers on their own 🥹 I suggest an experiment and she immediately shows the corresponding table 🥹🥹 some are even helping supervise interns 😊
so this is how it feels when you can hire PhD students with 2-3 papers already.
I could get used to it 😎
0
2
82
today after my talk, an undergrad came and told me: “you know most talks are like a gov person speaking, your talk was fun, confident, and engaging. it was very nice listening to you.”
can we tell this to ftm from 7 years ago who cried because of harsh criticism after a talk? 🥹
4
1
83
we replace the encoder of SimpleBEV with DINOv2 + LoRA and
- it becomes more robust to various corruptions as we show on nuScenes-C (nuScenes with corruptions).
- we can significantly reduce the encoder parameters that need to be trained (37M in ResNet101 vs 1 or 3M with DINO)
2
18
83
hiking in the northeastern part of Turkey with my sister, it has been amazing so far. highly recommended!
2
0
81
got invited to an interview on national TV. first, I thought cool, I get to say hi to my mom on TV 😄
but as it gets closer -it’s tomorrow- I’m like f-ck 😱 how do you explain stable diffusion or 3D vision to general public? questions are that specific!
6
0
82
ECCV reviews are coming, tough girl day 💪
me: I’m not gonna be result-oriented, enjoy the process and stuff.
also me: I’m gonna die if all reject again 😭
6
1
80
got an outstanding reviewer award at ICCV, back in the game! I think I always do the same thing but some conferences, I don’t get it, some, I do. I don’t know why, I guess it also depends on the AC receiving the review.
2
0
79
Nursena and Caner’s work on learning locally rigid registration for structural deformation monitoring has been accepted to the Differentiable 3D Vision and Graphics workshop at ICCV 2021. two undergrads 😍 my first paper with undergrads! in collaboration with Sinan Açıkgöz.
4
3
77
Our
#3DV
paper “Self-Supervised Monocular Scene Decomposition and Depth Estimation”
We jointly segment moving objects and model their motion separately in addition to egomotion to improve monodepth2, especially on Cityscapes & DDAD.
3
9
76
whole week, I met ppl from industry (friends or cold DM) and asked what my students need to do to get internships, PhD fellowships, or future jobs from them because we’ve been getting rejected a lot. I don’t have clear answers yet. if you have any advice on that, feel free to DM.
2
2
75
do you realize what’s happening in TR, the disaster that hit us? tens of thousands of ppl are dead, hundreds of thousands are out in the cold, afraid of going in their homes because of aftershocks. we have more help now but tonight is probably the last chance for rescue efforts.
7
9
73
tam vazgeçicem, onların hevesini, azmini görünce kendime geliyorum. eskiden çalıştığımın en az 3-5 katı çalışıyorum ama iş tatmini, motivasyon 10 kat. bayramınız kutlu olsun gençler! sizi geri tutanları, moral bozanları çok takmayın, yarınlar sizin :)
1
1
73
why would you care how a randomly initialized network reacts to your observations? because it could help you detect novel samples and then do a better exploration.
interesting findings with PPO + uncertainty-guided exploration.
2
9
73
don't you love twitter after conference decisions? well of course we'd love to hear about your x/y papers getting accepted when our x=y=1 paper, our precious, got rejected. started as Smeagol, finished as Gollum in the dark caves of research 😂
4
0
70
soo CVPR frenzy is over, I'm very proud of our work. we did our best and got very good results but most importantly, we openly admit our limitations. if you're gonna review our paper, you need to do better than repeating the weaknesses that I list at the end. I love writing about
1
3
70
"Veri Biliminde Kadınlar Konferansı" bugün!
1'de başlayacak aşağıdaki konuşmaları bu linkten izleyebilirsiniz:
benim konuşmam da 2'de!
(doçent diyor ama 🙈 o benim)
#WiDS2021
#WiDSIst2021
0
12
67
AI Lab olarak hemşirelik ile buluştuk. kanserli çocuklarda acının otomatik tespiti, hastayı rahatlatan ses tonu öğrenilmesi, kadına şiddette erken tespit için risk değerlendirmesi gibi konulardan bahsettiler. yaptığım her şeyi bırakıp, my sword is yours diye diz çökmek üzereyim.
1
1
69
3 of my students graduated from Masters as of today. I’m very proud of the work they have done, also with the help of our collaborators. Kaan has papers at ICCV and ECCV. Sadra has a 3DV paper and a BMVC submission, to be rejected 😅 Cagan’s seems to have a better chance at BMVC.
2
1
68
yayy our EuroHPC proposal got accepted 🥳
we are not GPU-poor anymore, like not GPU-rich but also not poor 😄
1
0
67
waiting for the ECCV results and my COVID test results. pretty sure that one of them will be positive, hoping for two positives 🤞🏽
3
0
66
it’s CVPR rebuttal week and there is also a project report due and ECCV is a little more than a month away, and I’m going to a spa hotel. what’s wrong, I don’t see a problem with that. in fact, that’s exactly when you need some relaxing, no?
4
1
65
so I'm not at ICCV but I was still able to bore ppl from afar 😅
thanks to the organizers of the workshops on uncertainty estimation () and trajectory forecasting ()
0
1
65
don't wait until graduation to apply to work with us. to give an example: Gorkay started working with us in his last year. in 6 months, we submitted a CoRL paper, it got rejected but with what we learned, we got an ICCV paper. he also has a NeurIPS submission. so apply now!
4
2
65
a student approaches me in the class:
- hi prof, I've heard so many things about you, I wanted to come and say hi.
me: oh, hi! good things I hope? 😅
- no, not really :/
well, at least, I evoke powerful emotions 🥲
2
0
63
somewhere a NeurIPS reviewer is rating our paper “not novel”, I can feel it. it’s like my knees hurting before it rains.
0
1
63
annemin kötü review’lere yorumu: “bırak iş kaybı olsun, baş kaybı olmasın da.” bizim ailede baseline ortaçağ :)
1
0
60
us: look at our cool method.
reviewer: but it’s not better than SOTA?
us: but our method gets similar results with a much more smaller model (guess why 🥹)
reviewer: so it is just a smaller version of the existing methods then?
us:
4
1
59