Skip to content
View MarkoHaralovic's full-sized avatar
💭
Training something.
💭
Training something.

Highlights

  • Pro

Block or report MarkoHaralovic

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
MarkoHaralovic/README.md

Hi 👋, I'm Marko

Computer Vision Researcher from Zagreb, doing Master Thesis at ETH in collaboration with Oxford.

  • 🌱 Currently exploring multimodal feature alignment, feature representation, open-vocab object detection, fairness, and VLM finetuning

Pinned Loading

  1. DL_cyber_attacks DL_cyber_attacks Public

    Implementation of Backdoor Attacks and Defenses on Resnet18 and EfficientNetB0. We have implemented Bad Nets and Data Poisoning attacks,along with Neural Cleanse,Fine-Pruning and Jittering defenses…

    Python 5

  2. geo-image-retrieval geo-image-retrieval Public

    Forked from onatakca/geo-image-retrieval

    Comparing linear evaluation and pooling methods of vision encoders (DINOv3, Perception Encoder, CLIP and StreetCLIP) on MSLS London dataset for image-to-image retrieval.

    Jupyter Notebook

  3. melanoma-classification melanoma-classification Public

    Fair and debiased skin lesion melanoma classification code. Using recall based cross-entropy loss combined with domain-discriminative training for equal perforamnce across all skin color groups.

    Jupyter Notebook

  4. gemma3_lora_ft_vs_icl gemma3_lora_ft_vs_icl Public

    Comparing Gemma3 models performance (270m,1b and 4b) on classification, reasoning and QA suite using ICL vs LoRA finetuning. Comparing performance to BERT trained on classification and QA.

    Jupyter Notebook

  5. CrisisKD CrisisKD Public

    Forked from onatakca/synchain-absa-emotion

    CrisisKD: Knowledge distillation in aspect-based sentiment and emotion analysis using Qwen2.5-32B as the teacher model and Qwen2.5-7B and LLaMA3-8B as student models.

    Python