diff --git a/heterogeneous_data/multivariate_models_v1_7.ipynb b/heterogeneous_data/multivariate_models_v1_7.ipynb
index 4f19b0070e3b868c54b43bdf6f2f82267cdf9f0a..f90ab02b012c1a31eb9f8f38d150b67089739116 100644
--- a/heterogeneous_data/multivariate_models_v1_7.ipynb
+++ b/heterogeneous_data/multivariate_models_v1_7.ipynb
@@ -1663,19 +1663,19 @@
     "\n",
     "Can you find the reason why we use $\\log$ values to parametrize the variance (W_logvar)? \n",
     "\n",
-    "<img src=\"img/vae.svg\" />\n",
+    "<img src=\"https://gitlab.inria.fr/epione/flhd/-/raw/master/heterogeneous_data/img/vae.svg\" alt=\"img/vae.svg\">\n",
     "\n",
     "## Sparse VAE\n",
     "To favour sparsity in the latent space we parametrize the variance of the latent distribution with log_alpha (times $\\mu^2)$.\n",
     "\n",
-    "<img src=\"img/sparse_vae.svg\" />\n",
+    "<img src=\"https://gitlab.inria.fr/epione/flhd/-/raw/master/heterogeneous_data/img/sparse_vae.svg\" alt=\"img/sparse_vae.svg\">\n",
     "\n",
     "## MCVAE\n",
     "The MultiChannel VAE is built by stacking multiple VAEs and allowing the decoding distributions to be computed from every input channel.\n",
     "\n",
     "Excercise: sketch the Sparse MCVAE. How many (log) alpha parameters do we need?\n",
     "\n",
-    "<img src=\"img/mcvae.svg\" />\n"
+    "<img src=\"https://gitlab.inria.fr/epione/flhd/-/raw/master/heterogeneous_data/img/mcvae.svg\" alt=\"img/mcvae.svg\">\n"
    ]
   },
   {