katja seeliger
@seelikat
Followers
966
Following
5K
Statuses
2K
Links between human and machine vision. #DeepLearning & #Neuroscience. PostDoc at ViCCo group by @martin_hebart at @MPI_CBS. Opinions are my own.
Leipzig, Koffer in Berlin
Joined July 2012
@DrBaffi_Jr @skdh @XPhyxer1 @NTFabiano They are unlikely to go down that route though (although massive profits would allow them to), but will instead start some ill-incentivized pay-by-review program. Probably time to look towards what LLMs can do for us instead š¬
1
0
0
RT @mtoneva1: @meenakshik93 Indeed! We in fact recently showed this in language--optimizing a speech language model to predict brain recordā¦
0
2
0
@danielbigham @BernardJBaars Hadn't encountered the topic before getting into neuroscience with a CS background. Had extreme materialist views. Now I think consciousness presents the biggest mystery. Doesn't even have a proper definition. Not in the research but have read broadly since realizing the problem.
1
0
2
RT @paolo_papale: šØšØšØ New Dataset out! Today we release the THINGS Ventral-stream Spiking Dataset (TVSD) to become part of @martin_hebart'ā¦
0
27
0
RT @AntonioLozanoDL: @paolo_papale et al.'s new vision dataset contains recordings of 1024 channels in V1, V4 and IT of 2 NHPs, and 22k+ imā¦
0
3
0
RT @lynnle_ai: Our new reconstruction paper using Inverse Receptive Field Attention (IRFA) is online! IRFA reconstructs images and visualā¦
0
8
0
This year's Algonauts challenge will be on the excellent @CNeuromod data.
The Algonauts Project 2025 Challenge: How the Human Brain Makes Sense of Multimodal Movies.
0
1
3
The mission to restore sight takes a big leap forward! š š
Cortical visual implant pre-print announcement! š Now out in @medrxivpreprint , together with Xing Chen, @Pieters_Tweet et al. Brain implant enterprises such as @neuralink and @Phosphoenix_BV aim to tackle some of the biggest challenges in sensory restoration, like #blindness. In the case of visual neuroprostheses, building a āphosphene mapā for each user will be critical to conveying useful visual information to the blind users (aka device calibration). This map matches the ID of each stimulated electrode in the visual cortex to the location of the perceived artificial visual percepts. Our experience with blind volunteers and monkeys shows that high-resolution implants with hundredths to thousands of electrodes can mean hours of tedious and sometimes imprecise or unreliable calibrations for future users. This is a problem! We solved this. Iām happy and proud to present NEUmap: NEural Unsupervised electrode mapping. Want to know more? ... 1/
1
3
7
RT @biorxiv_neursci: Brain-Guided Convolutional Neural Networks Reveal Task-Specific Representations in Scene Processing
0
4
0
RT @SakanaAILabs: Introducing ASAL: Automating the Search for Artificial Life with Foundation Models Artificial Lā¦
0
645
0
RT @seelikat: Reconstructions of seen image stimuli from brain activity on the diverse THINGS test set. #neurips2024 #neurips2024 This metā¦
0
3
0
Don't miss out on Lynns awesome work on mind reading at #NeurIPS #NeurIPS2024 ! Real-deal visual reconstruction, no diffusion models :)
Today at 11:00-14:00, Iāll present our work on neural decoding: reconstructing images from brain activity using convolutional neural networks and brain-inspired theory to improve model interpretability. Join me in East Exhibit Hall A-C, poster #3808!
0
0
1
RT @patrickmineault: New post! What do brain scores teach us about brains? Does accounting for variance in the brain mean that an ANN is brā¦
0
68
0
RT @patrickmineault: Excited to release what weāve been working on at Amaranth Foundation, our latest whitepaper, NeuroAI for AI safety! Aā¦
0
95
0
RT @Dr_Alex_Crimi: New #brain modality: #EEG on scalp š Temporary tattoo printed directly on the scalp offers easy, hair-friendly solutionā¦
0
26
0
Reconstructions of seen image stimuli from brain activity on the diverse THINGS test set. #neurips2024 #neurips2024 This method does not use a text-to-image diffusion model ( see here: ). Review discussion:
Thank you to my coauthors: @paolo_papale @seelikat @AntonioLozanoDL @ThirzaDado @Pieters_Tweet @marcelge @yagmurgucluturk @umuguc Full set of reconstructions:
1
3
12
RT @seeingwithsound: #MonkeySee: Space-time-resolved reconstructions of natural images from macaque multi-unit activity
0
5
0