Sign In

Flowers test sdxl - flowers test sdxl Showcase

/7DD9C97CA4372C83F1B17C59295F86DF4A872479AC97432891A23DD1DCA441C9
AF4AAAEBDCCB69D27B2AA616476AFCBC8A13A55064052F1F84C3B34B816C40B3.jpg
/F490D63975C6910F3139FBCC347DD8C8C001D21EB33D0C5055307D45B03ABD40

Project Synopsis:

  • Experiment Overview:

    • Conducting a multifaceted test of LoRa's capabilities in AI image generation.

    • Three phases: Onsite Generator, Auto1111, and Comfyui exploration.

  • Phase 1: Onsite Generator with Checkpoints

    • Trained LoRa model with 100 flower images across 10 diverse checkpoints.

    • Utilized onsite generator exclusively to observe LoRa's adaptation and image generation.

    • Aimed to establish baseline performance in a controlled environment.

  • Phase 2: Auto1111 Exploration

    • Replicated Phase 1 parameters on Auto1111 platform for comparison.

    • Evaluated LoRa's performance and output quality in a different AI environment.

  • Phase 3: Testing with Comfyui

    • Conducted Phase 1 parameters on Comfyui platform to further explore LoRa's capabilities.

    • Assessed how different AI platforms influence image generation outcomes.

LoRa Details:

  • LoRa model trained with 1000 flower images.

  • No captions provided during training.

  • Epoch set to 14.

  • Hypothesis: Training LoRa with diverse flower images will produce usable outputs.

  • Sharing LoRa model and dataset for public access and collaboration.

Please post all pictures created with this lora to the gallery directly.

Comments